On the (Non)complementarity of the [theta]-Theory and Checking Theory

Abstract

Chomsky (1995) proposes that the θ-system and the checking system form two complementary modules. As a consequence both subjects and objects must form nontrivial chains to check their formal features with a functional category (T and v, respectively). I argue that objects and exceptional-Case-marking subjects check their formal features with a lexical verb, whose domain is therefore both θ-role assigning and feature checking. I show that discarding the complementarity assumption in this manner results in a more "bare" theory of the computational system as well as several empirical advantages.

Chomsky (1995) observes that although θ-theory should be part of any language in the broad sense, checking theory is part of the language faculty only and therefore in need of explanation. He then claims that these two modules are complementary:

The displacement property reflects the disparity-in fact, complementarity-between morphology (checking of features) and θ-theory (assignment of semantic roles), an apparent fact about natural language that is increasingly highlighted as we progress toward minimalist objectives.

(p. 222)

I agree with Chomsky that displacement is a fact of natural language and that something like the Attract/Move operation can account for it. However, Chomsky's claim that the θ-system and the checking system are complementary is not a fact; rather, it is an assumption of the framework used, an assumption that I argue can be advantageously dispensed with. As Bouchard (1997) notes, in the LGB framework (Chomsky 1981) objects are assigned Case in the same position where they are assigned a θ-role, which can hardly be the case if complementarity is a fact.

Chomsky's complementarity assumption leads to two interrelated consequences: θ-positions and checking positions are two nonoverlapping sets, and arguments must form nontrivial chains, each with a foot that receives a θ-role and a head that checks its formal features against a functional head that has an array of matching uninterpretable features. Thus, feature checking is an additional edifice built on top of a θ-role assignment foundation. I would like to propose that θ-role assignment and feature checking are not as neatly separate within the computational system as this theory represents them to be. As a matter of fact, I argue that they overlap partially: the operation Attract/Move, which is brought in by the need to check features, may have consequences for the θ-structure of a sentence; also, Merge of arguments, which is triggered by θ-role assignment, may give rise to feature checking as a by-product. Thus, I claim that once the checking system is built into the language faculty, there is no reason to assume that it could not intermingle with the θ-system. I argue that it does so and therefore that our model of syntax should reflect this fact. I show that if we change Chomsky's assumptions, we obtain a more streamlined theory of syntax and a more accurate picture of how the computational system works.

As a consequence of my argumentation, I will conclude that nominative Case checking and accusative Case checking do not work along parallel lines, contrary to what has been the majority opinion among researchers in this area for the past decade. In particular, Chomsky proposes that objects form nontrivial chains to check their formal features, including accusative Case. Further, the head that checks these features is the same one that introduces the external argument. Instead, I argue that objects are fully licensed by the lexical verb that assigns them a θ-role and that ECM (exceptional-Case-marking) subjects are attracted (or pied-piped) by this lexical verb.

This article is organized as follows. In section 1 I argue that Chomsky's proposal concerning accusative Case checking leads to an extra stipulation, perhaps two, that restrict θ-role assignment and feature checking and detract from the general elegance of the theory. In section 2 I propose that the features of the object are checked by a lexical verb, thus reducing the light verb to the role of introducing the external argument and allowing the above...