This dissertation presents a theoretical analysis of the interaction between the clauses in correlative constructions in Hittite at the syntactic, semantic, and discourse level. I argue that the relative clause is not connected with the main clause in the syntax, only at the discourse level. I defend this claim by examining the syntactic and semantic relationships that the correlative has with the main clause and the resumptive correlate.
I argue that the correlate is a discourse anaphor coreferent with the correlative, not a variable bound by it. This is the simplest explanation of the fact that the distribution of NP types as correlates is completely explained by Hittite-wide principles governing the distribution of NPs as discourse anaphors. There are no special requirements attributable to the correlative construction itself. Moreover, numerous correlatives are linked indirectly to the main clause and not resumed by a coreferent correlate — a fact incompatible with variable binding but ordinary for discourse anaphora.
I argue that the correlative’s position cannot be derived by movement from within the main clause. Moreover, the correlative is not syntactically integrated into the main clause, despite being semantically dependent on it. The correlative is a clausal hanging topic and is external to the main clause, linked to it only in the discourse. This accounts for a variety of complexmulti-clausal correlative constructions that pose difficulties for integrative approaches.
I propose a model of correlative semantics framed in Segmented Discourse Representation Theory (Asher and Lascarides 2003), a dynamic framework that models the rhetorical relationships between segments of discourse. I posit a function REF that makes the correlative into a referential expression and a rhetorical relation HT that predicates the main clause conditionson the correlative’s referent(s). I demonstrate how this model encodes the characteristic maximal interpretation of correlatives in definite and indefinite readings as a reflection of referent identifiability in context.
A common assumption in the theoretical literature is that dependent clauses are syntactically subordinate to their main clause. I articulate a different view of the division of labor between syntax and discourse, and I suggest that standard assumptions should be re-examined.