Hittite correlatives are paratactic

Abstract

This paper argues that correlative constructions in Hittite are paratactically structured. The relative clause is essentially a clausal hanging topic, sitting at the left edge of the main clause in linear juxtaposition without actually being an integrated part of it, syntactically speaking. I defend this claim in two stages. First, I argue that correlatives in Hittite are base-generated in their left-edge position rather than derived through movement (as advocated for Hindi by, e.g., Bhatt 2003). I adduce as evidence the fact that the main clause correlate appears to be simply a discourse anaphor and need not even be present in the construction; these observations are incompatible with a movement-based derivation that generates the relative clause as a modifier of the correlate. There is also evidence for a lack of locality effects. The second part of my claim, that Hittite correlatives are not syntactically integrated, differs from most base-generation accounts of correlatives, which take correlatives to be clausal adjuncts. I support my position with parallels to hanging topics and peripheral adverbials (Haegeman 2012) and with examples of intervening non-subordinate clauses, and I offer some comments on why syntactic adjunction is less well suited to the Hittite situation. I also show that Segmented Discourse Representation Theory (Asher & Lascarides 2003) accommodates these structures, including some notably non- canonical ones, in a simple and principled way.