Long-Distance Agreement and the Syntax of for-to Infinitives

Throughout the history of generative grammar, expletive constructions in English and other languages have been a major focus of interest. The long-distance agreement effect attested in such constructions (see (1a-b)) has always represented a major theoretical challenge, a challenge that motivated some of the innovations of recent work within the Minimalist Program. Chomsky(2000), for instance, suggests the operation Agree as part of his account of this effect.

(1)

1. a. There was/*were elected an unpopular candidate.

2. b. There seems/*seem to have been elected an unpopular candidate.

Avoiding unnecessary technical details, the main points of Chomsky's proposal are the following. The long-distance agreement effect attested in (1a-b) is due to a "T-associate relation that involves features only and is independent of the expletive" (p. 126). This relation involves a probe (T) and a goal (the associate, an unpopular candidate) within the domain of the probe (the sister of T in this case) and effects the erasure of uninterpretable features of probe and goal. This operation is referred to as Agree. The satisfaction of Case-theoretic requirements of the associate (i.e., the erasure of an uninterpretable Case feature) is one effect of this operation. As for expletive there, because of its formal properties, this element has no need for structural Case. It merges with T, thus becoming Spec, T, and in this way satisfies the EPP feature of T.

For-to infinitives such as the bracketed parts in (2a-b) represent a difficulty for this approach.

(2)

1. a. [For there to be a unicorn in the garden] would be a surprise.

2. b. It is unimaginable [for there to be a unicorn in the garden].

The difficulty is this. If a (long-distance) T-associate (probe-goal) relation is the mechanism by which post verbal subjects are licensed, one does not expect (2a-b) to be grammatical given the general fact that an infinitival T is not the right element to be engaged in such a relation with an overt, lexical NP/DP. As is well known, a probe-goal relation with an infinitival T could never license an overt preverbal subject in infinitival clauses.

(3) *It is unimaginable [Mary to arrive on time].

This reasoning is supported by the ungrammaticality of (4a-b). These examples, to be contrasted with (2a-b), show that the occurrence of for in such infinitives is crucial.

(4)

1. a. *[There to be a unicorn in the garden] would be a surprise.

2. b. *It is unimaginable [there to be a unicorn in the garden].

Note that there in these constructions would not need to enter into any relation with for in order to be licensed since, by hypothesis, it does not need to be involved in any relation of Case assignment/ checking. As seen above, there may occur in finite sentences such as (1a-b) where, according to the theory under consideration, T is (directly) engaged in an Agree relationship with a postverbal subject whereas there serves simply to satisfy (check) the EPP feature of T. Thus, within the approach under consideration, the ungrammaticality of (4a-b) could not be due to a failure to satisfy some formal requirement of there.

Given examples like (2a-b) and (4a-b), an attempt to maintain an approach based on Agree leads inevitably to an assumption that grammatical cases like (2a-b) involve a long-distance probe-goal relationship between the postverbal subject and for (rather than T). If an assumption along these lines turns out to be wrong, a different account of the syntax of postverbal subjects in examples like (2a-b) (and generally) must be sought, given that T and for are, under standard assumptions (but see discussion below), the only apparent candidates for entering into an Agree relation with a postverbal subject in such cases.

The crucial question is then, how does for fit into the syntax of for-to infinitives so as to render sentences like (2a-b) grammatical? The postulation of a long-distance probe-goal relationship between for and a postverbal subject is easily refuted. Emonds (1985:297) observes that sentence...