This dissertation, presented before a committee including sociolinguistic eminence grise William Labov, has broad goals, attempting to illuminate the ways in which sociolinguistic work on variation can benefit from understanding of theoretical developments in phonology, specifically the then-recent first go-round at optimality theory from Smolensky et al., with all their PARSE and FILL and whatever. It's too harsh to call that terminology "outmoded", but it has been superseded to a degree, which for me at least aggravated my natural tendency to skip straight to the executive summary of the OT analyses and not worry about the tableaux and shit. Like, just give me the take-away! no doubt that will come back to bite me.

On that broad level, though, I admire Reynolds's commitment to nailing down exactly what variation IS in a way that is more satisfying than what phonologists have traditionally gone forward with, and taking the complications as opportunity rather than crisis. Indeed, when more than one form of a word is optimal, is that because speakers are participating in multiple grammars, or because the constraints that evidently leave us with more than one form are unrankable, the last-form-but-one uneliminatable?

Reynolds answers the question deftly by conceding the existence of multiple grammars in the case of consistent dialectal variation, be it stable or new-forming, and in the case of variations that are not consistent intra-speech community (or intra-speaker), he introduces the concept of the "floating constraint"--alongside but unranked relative to a language's normal constraint chain, and then when you can tie it down to a specific position with a specific surface form, well, there's your difference between one dialect and another, right there! So the more you fill out the constraint ranking of a language, the more you can say "take "standard" English, move this over here and you've got Brummie or Nuyorican; move these twelve or two hundred like this and you've got Korean!" So economical, so pleasing.

I have to admit I was hoping he would conclude that sociological causes of variation could be explained by OT, resulting in constraints like *COMPLEXCODA-CASUAL to get alternation between, like, "thinking" and "thinkin'" (although I'm aware that the latter is a vestigial progressive suffix /-end/) or DEL-NUCLEUS[-stress]-INGROUP to explain my uncle Walter's "hos khnur khop?", or ridiculous excesses like DOUBLE-C[perf] to get that yakuza-movie "-yagarrrrrre!", but the framework can only be made to do what the framework can do. Congratulations, Dr. Reynolds! ( )