Clause Structure and Null Subjects: Referential Dependencies in Korean

Abstract

This dissertation investigates control structures with infinitival and jussive complements in Korean, an empirical domain that has so far received little attention. It focuses on three theoretical issues: the licensing conditions for controlled subjects, their categorial status, and the algorithm that determines the choice of an argument that acts as a controller. The results show that controlled subjects are not a homogenous class, even within a single language, and that their categorial status correlates with the structure of the complement clause in which they occur.

I examine control constructions with complements introduced by the complementizers -kilo, -lyeko, -koca and -tolok. I propose new diagnostics to demonstrate that they are infinitival clauses, and argue that the forward and backward control dependencies they exhibit result from A-movement of the controller followed by chain linearization at PF. The subject of these infinitival complements is thus the copy of an argument that has raised. Moreover, I compare tolok-complements with Raising-to-Object constructions, as the two share a number of properties, like allowing the controller or lower subject, respectively, to exhibit Case alternations. I make the novel observation that the parallelism in their behavior ends when their complement clauses are fronted by scrambling, and argue that such an asymmetry is explained by a derivational model that adopts Cyclic Spell-Out.

I also examine control constructions where the complement is a jussive clause. Jussive clauses comprise imperatives, exhortatives and the (cross-linguistically rare) type of promissives. Interestingly, in Korean they can be embedded and, when embedded, they exhibit different control patterns. I show that the null subject of jussive complements is not the copy of a raised argument, but rather a null pronominal element bound by a discourse participant operator in the left periphery. I argue that control dependencies in this case result from a shift in indexicality of the embedded subject pronoun, a phenomenon available in many languages. In particular, I propose that the range of operators available in the jussive complements is determined by the class of control predicates, and furthermore that they are connected to an argument in the matrix clause by syntactic binding.