Long-Distance Agreement in Tsez: A Reappraisal

This paper is a critique of Agree-based accounts of long-distance agreement in Tsez. It highlights the inconsistencies inherent in such analyses and suggests instead an alternative where agreement is accomplished in ‘local’ configurations brought forth by the structure-building operation Merge and Move/Remerge. Contra what has been proposed before, the paper suggests that the agreement triggering nominal has an uninterpretable structural Case feature checked as a reflex of phi-feature checking with the matrix verb. Introduction Tsez, a Nakh-Dakhestanian language spoken in the northeast Caucasus presents an intriguing example of phi-feature agreement on verbal heads triggered by nominals across (apparently) finite clauses. Recent studies (cf. Bhatt, 2005; Polinsky and Potsdam, 2001) try to incorporate the phenomenon within Chomsky’s (2000, 2001a, b) long-distance Agree based paradigm, wherein agreement is realized without inducing movement of the trigger to the probe’s ‘checking domain’ (in the sense of Chomsky, 1993). This paper exposes some crucial problems in these studies and suggests a ‘local’ alternative. The core assumption of my analysis is that the agreement triggering (absolutive) nominal adjoins to the embedded TP, which in turn pied-pipes to the specifier of matrix v. By virtue of being adjoined to the moved phrase – the element is also within the ‘checking domain’ of the matrix verb (also see Frank, 2005; Koopman, 2003). However, this paper provides another independent solution, which rationalizes the idea that elements adjoined to specifiers – at least in some instances – may trigger agreement with the respective head. I argue that adjunction to TP is a step of agnostic movement (Boskovic, 2002; Franks and Lavine, 2006) of the absolutive in an attempt to remain as close to v as another potential goal, the TP itself. Despite this step of movement, the TP being the closest goal – moves and merges with v. The absolutive subsequently reprojects (à la Hornstein and Uriagereka, 2002). This makes the adjoined element sister to the target. ‘Longdistance-agreement’ between v and the reprojected DP thereby ensues in a ‘local’ configuration. My analysis also provides some evidence calling for a stricter alliance between case and agreement checking. I maintain that absolutive case on 1 I say ‘apparently’ because these clauses cannot take overt complementizers (more on this below). 2 For more elaborate discussion on the constraints on reprojection and effects on linearization and scope, see Chandra (2006).