Aspects of ergativity

This dissertation examines two kinds of split ergativity found in Georgian, the Caucasian language of the Republic of Georgia. First, certain Georgian clauses have an ergative case-marking system but an accusative verbal agreement system, so that the case and agreement systems 'disagree'. Second, the case-marking and agreement patterns in Georgian clauses vary according to the tense and aspectual properties of the events denoted by those clauses. These phenomena are problematic for current theories of case and agreement marking couched in terms of Chomsky's Minimalist Program, because in that framework, (i) case and agreement are taken to be checked together in a single projection, typically AgrS or AgrO, and hence should always 'agree', and (ii) case and agreement are not explicitly tied in any way to tense/aspectual features. In this dissertation a theory of case and agreement checking is proposed within the framework of the Minimalist Program which can account for the phenomena found in Georgian. The proposal has two facets. First, a theory of A-movement is developed in which A-movement is subject to a minimal link condition but is in principle unbounded. Evidence for this view is adduced from ergative languages and from grammatical cases of superraising found across languages. Second, a theory of clause structure is developed based on the hypothesis that case and agreement are reflexes of operations whose purpose is to check aspectual features of the clause. Under this view, Agreement Phrases are taken to be Aspect Phrases, and A-movement is taken to check aspectual features borne by NP arguments against corresponding aspectual features of the Aspect Phrases. Evidence for this view of clause structure is adduced from syntactic and morphological facts from Georgian. In particular, it is shown that the four major morphological classes of Georgian verbs each correspond to different aspectual types of events, each of which projects a different arrangement of Aspect Phrases and commands a different case/agreement paradigm. The theory is then shown to correctly predict the case and agreement facts in Georgian.