Forum Lecture

Mark Baker (Rutgers University)

"Formal generative typology: A little vision and a little example"

Thursday, 23 July 2009, 7:30 pm
2050 Valley Life Sciences Building

I begin by outlining why I think there is both opportunity and need for a sort of typological research that is not so commonly done. On the one hand, it borrows from Chomskyan-style linguistics a willingness to use abstract linguistic representations in order to capture substantive linguistic generalizations that diverse surface forms have in common. On the other hand, it borrows from Greenbergian-style linguistics a commitment to evaluating putative linguistic universals over a genetically and areally diverse sample of languages from the beginning. This challenging pair of commitments can be made more feasible by being applied in the first instance to an intermediate number of languages — an order of magnitude more than many generative studies, but an order of magnitude less than some current typological studies.

I then illustrate the advantages of this approach more fully by presenting a fairly modest case study — a crosslinguistic comparison of agentive nominalizations (the finder of the wallet) and event nominalizations (the finding of the wallet). It is well known that event-denoting nominalizations can have verb-like internal structure in various ways (finding the wallet so quickly) across a wide range of languages (Koptjevskaja-Tamm 1993). What is generally not appreciated is that agent-denoting nominals cannot have similar verb-like internal structure (*the finder the wallet so quickly) (but see Comrie and Thompson 1985 for an early observation). I conjecture, in fact, that agent-denoting nominalizations of this sort are universally impossible, for reasons having to do with how the semantics of agency interact with the nature of nominalization. Along the way, I show how Chomskyan techniques of inquiry into particular languages (English, Sakha, Mapudungun) allow one to sharpen the hypothesis in crucial ways; this then provides the basis for Greenberg-style comparison of unrelated languages. Crucially, what is crosslinguistically ruled out is a particular kind of syntactic structure — not the function of referring to the agent of an event while still having that event modified by adverbs. (The latter function is perfectly possible if the structure used is a headless relative clause.) Thus, a method that is relevantly like Formal Generative Typology is probably crucial to discovering this universal feature of language (if it is one).