Gribanova colloquium

January 24
Rosenwald 011
Stanford University
Case and agreement in Uzbek nominalized clauses

In this talk I use novel evidence from Uzbek nominalized clauses to shed light on the nature of the connection between case licensing and agreement, and to distinguish between two competing approaches to this connection. Baker and Vinokurova (2010) have argued that two modalities of case licensing are necessary to account for the entire range of case patterns in Sakha. One modality involves case licensing via a structural link (AGREE) between the relevant noun phrase and a functional head, which simultaneously values phi and case features (Chomsky, 2000, 2001). A second is configurational case assignment (Marantz, 1991), in which a noun phrase is assigned case on the basis of the head that c-selects it or on the basis of that phrase’s position with respect to other noun phrases in the same clause. Their argument for case assignment via AGREE with a functional head rests crucially on evidence from Sakha nominalized clauses, in which structural subjects receive genitive case marking: genitive on subjects is licensed if and only if there is also agreement. Levin and Preminger (under review) have argued against the position that such evidence forces the use of the functional head case licensing approach, and provide a purely configurational account of the same set of facts.

New evidence from case-marking on structural subjects in Uzbek not only supports the latter position, but also leads to a further step: I argue that the Uzbek facts are actually incompatible with an account in which the same functional head is responsible simultaneously for establishing agreement and case relations. I use evidence from three types of Uzbek nominalized clauses, all of which converges on the idea that structural case licensing and agreement are controlled by fundamentally distinct syntactic mechanisms. The case of structural subjects in Uzbek nominalized environments is either genitive or bare/nominative; both obligatorily trigger agreement. Though the choice of case for subjects appears optional on the surface, I demonstrate that there are positional differences between subjects marked genitive and those marked nominative/bare, the former being structurally higher than the latter. I provide evidence that this higher position is an A-position to which any argument, including subjects, can be short-scrambled. What is striking is that the link in the presence of genitive on subjects of nominalized clauses and the presence of agreement is easily disrupted by word order manipulations: when a non-subject argument is scrambled across the subject, this subject may be marked only nominative/bare, never genitive (presumably because it must stay low). Despite the lack of genitive case assignment, though, agreement is nevertheless required. Since my claim is that the case marking pattern and the agreement pattern are controlled by distinct mechanisms, I discuss each mechanism in turn, developing the idea that the Uzbek facts are best accounted for by a theory which de-links case and agreement to a significant degree: case is relevant for agreement, but not the other way around.