Thematic roles and argument structure
This course will follow the syntactic development of the empirical understanding of argument structure and of the related theoretical device of thematic or theta-roles. We will begin with some background on what thematic (or theta) roles are, how they fit into grammatical description and what the points of controversy are in their definition and use. We’ll then look in detail at the specific theoretical device of theta-roles and the classic theta criterion of GB theory, see how it was motivated as independent of case and (structurally defined) grammatical functions, look at its implications for different types of A-movement and for the control vs. raising debate, and consider its main strengths and weaknesses. Then we’ll discuss different proposals about how to implement or extend theta-roles and the theta criterion, in terms of more fine-grained structural positions(Baker’s UTAH, Hale & Keyser), in terms of syntactic heads dedicated to introducing particular argument types (Kratzer, Pylkkänen, Lohndal), or by rethinking theta-roles as syntactic features assigned in the course of the derivation (Hornstein).
Following this we will consider ways of thinking about thematic roles that depart in one way or another from the classic theta criterion, e.g. work deconstructing thematic rolesinto semantic primitives (Dowty) or countenancing the possibility that a single NP could receive more than one role (Hornstein, Ramchand, with precedents) and the various ways (thematic raising vs. something else) to implement the latter notion.
Finally, we will discuss and propose some ideas about what the theta criterion should be replaced with if we are to give up the one-to-one mapping between theta-roles and NPs. If an NP can in fact get more than one theta-role, how do we prevent the system from over-generating? What are the restrictions that do obtain to prevent such over-generation, and where in the grammar do they apply? Could a version of the strict theta criterion be maintained that is relativized to different types of roles, e.g. narrowly thematic roles vs. discourse-related roles? Relatedly, what kind of principle should replace the theta criterion in blocking a single NP from getting both PATIENT and AGENT roles with obligatorily transitive verbs, i.e. ensuring that “*Doris smacked” with the meaning of “Doris smacked herself” is impossible?