Joan Bybee

Introducing Exemplar Theory! (I refuse to call it "Exemplar Dynamics", because I am twelve, and the acronym is embarrassing.) I get how Optimality Theory is pleasing to a mind that enjoys systems, and it has a certain efficient thoroughness of desciption that ET, with its incremental changes and word clouds, certainly doesn't, but it was always artificial--a schemata applied after the fact, a reverse-written equation to get a result that makes sense. Good as a descriptive tool, but it didn't feel like the way language acquisition works.


And this does. Clouds of forms, and the more tokens you hear, the more you home in on a production target. I'm watching it happen with Luisa, who has just about definitively added the rhotic to /ma:din/,which makes it pretty damn close to my name really.

And the lenition of frequent forms as evidence--why else do you say /evri/, but /artIlɜri/? (Pardon the approximate IPA--I am lazy.) And the way it collapses lexicon and grammar, performance and competence--so organic, so anti-Chomsky. And Pierrehumbert, for it is her article I am using, even suggests the application of sociolinguistic tags to formants, which is super useful for my glottalization-in-Vancouver-women project. And there is no simple way to model this stuff, really, incremental as it is, and that's a flaw, but it also probably makes more sense and presumes less than generative models.… (more)