Heidi Harley (University of Arizona)
The colloidal connection between theta roles and grammatical
roles has been a preoccupation of generative grammar since the invention
of the former. Since grammatical roles are essentially an epiphenomenon of
structural position, the general direction was to propose that the one is correlated
with the other, culminating in the proposal of principles like Perlmutter's UAH
and Baker's UTAH, and the entire field of inquiry called Linking Theory.
Theta roles as explanatory devices, however, are decidedly unsatisfactory; as
solid as they seem initially, when examined in detail they are much harder to
pin down. Only when viewed at a sufficient level of abstraction can general properties
be discerned, such as the fact that there seem really to be just three theta-roles,
Agent, Theme and Goal (Baker 1997). This led Hale and Keyser to their notion of
syntactic
argument structure: 'theta roles' are an effect of structural position too.
Roughly speaking, H&K propose that theta roles are an epiphenomenon of
structural position in VP, just as grammatical roles are epiphenomenal to
position in TP. (To add that discourse roles are probably an epiphenomenon of
structural position in CP is very tempting at this point, but way beyond the
purview of this talk). Mismatches between theta role and grammatical function will
then be the result of movement of the DPs, or alteration of the VP structure by
various kinds of morphsyntactic processes during projection, or, most often,
both. The peculiarly fluid-yet-solid nature of possible mismatches is the
result of the strict restrictions on movement processes in combination with the
available alternative starting points provided by the morphosyntax.
I'll consider the development of Hale and Keyser's ideas over the course of the
past decade, and touch on their relation to (a subset of) the following topics:
how syntaxy is l-syntax? What's little v for? What does the semanticization
of H&K's ideas entail, and how is it different from generative
semantics? How about neo-Davidsonian semantics? What's more Minimalist,
eliminating OC PRO with theta features, or keeping OC PRO
and eliminating theta theory? What is the connection to category? Why is this
kind of
syntax not productive, that is, what is the domain of the Word?
Data-wise, resultative constructions, English structural paraphrases, and
dative shift will
certainly show up. Plus I will present three reasons for not NOT analyzing
"kill" as
"make dead".