Licensing Objects with and without Movement

The goal of this contribution is to determine whether object licensing requires movement or not. In a series of recent papers, Susi Wurmbrand 2001, 2003, 2004a, 2004b (see also Bobaljik & Wurmbrand 2003, 2004) has argued, mainly on the basis of German data, that we ought to reject the view that internal arguments universally move (either overtly or covertly) to SpecAgrOP (cf. Chomsky 1993, Adger 1994). Her proposal is that Agree (Chomsky 1998, 2000) is the optimal way of analysing accusative case licensing. A large part of her argumentation consists in showing that German direct objects do not move for case reasons, not even at LF. Hence, it is necessary to have an operation like Agree, which does not require movement in and of itself. This paper reaches a slightly different conclusion: I argue that Wurmbrand is right in suggesting that Agree is the way in which objects are licensed in West Germanic languages. However, I also argue that other languages require object movement to a VP external position. We compare two languages – namely, Basque and Dutch-, showing that a number of asymmetries between them fall out if object licensing involves movement in Basque, but not in Dutch. The choice between licensing objects via Agree or via movement is taken to be a point of parametric variation. This paper is organised as follows: section 2 we introduce three object-related asymmetries between Dutch and Basque, with brief references to other languages. In section 3, we explore how these differences receive a uniform account under the hypothesis that Basque has obligatory object movement, but Dutch hasn’t. Finally, in section 4, we explore some consequences of this hypothesis for the larger theory of word order and phrase structure.