On the Implementation of A -Movement
One salient property that intermediate positions have in common is that they normally cannot function as final positions. Rather, they are gates, or escape hatches, through which elements must pass to reach higher positions.
The paradox of these intermediate positions is that, on the one hand, they must autonomously cause a movement step (if we want to take seriously the idea that each step is locally determined, with no ‘‘look-ahead’’ to subsequent derivational steps), yet, on the other hand, we must ensure that movement will not stop there.
Let us make more specific assumptions about the implementation of movement in general, and then return to intermediate movement. I will follow standard minimalist assumptions about the fact that movement is guided by a system of morphosyntactic features: basically, the at tractor and the attractee are characterized by a certain shared feature specification.
In the system devised for A-chains in Chomsky 2000, uninterpretable features on the attractor and attractee play a crucial role: uninterpretable f-features on T enter into an Agree relation with a DP made active by uninterpretable Case features, thus establishing the configuration that is a prerequisite for movement.
(1)
Now consider A ¯-chains. In questions, for instance, the question feature Q, expressed by special complementizers like if in English and si in French, and by question particles in various languages (sometimes limited to yes/ no questions, sometimes generalized to all questions), is presumably interpretable; it enters into clause typing (Cheng 1991) and selection from a higher predicate. The morphosyntactic feature characterizing a certain class of operators as question operators (wh- in English, qu- in French, etc.) is also visible to, and used by, interpretive systems. So, the minimal morphosyntactic feature system required by questions (and other criterial constructions) does not seem to involve morphologically overt uninterpretable features.
We thus seem to have two theoretical options. Either we can assume abstract uninterpretable features associated to the interpretable features of attractor and attractee, making the system fully parallel to the A system in Chomsky 2000 (much as in Pesetsky and Torrego’s (2001) approach). Or we can take the lack of direct morphosyntactic evidence for A ¯-uninterpretable features at face value and assume the criterial at traction is achieved directly by the interpretable features; that is, in con figurations like (2), C attracts XP bearing a matching criterial feature.
(2)
Not much hinges on this point in the present context. The critical need for the computational system to work is to formally identify the attractor and the attractee, and this is achieved in both ways. For concreteness, in what follows I will take the second path, which minimizes the feature specification to be postulated; I will therefore assume that interpretable criterial features can serve as attractors of movement.
Whatever assumption one makes about attraction to a criterial position, the issue of uninterpretability arises in a different form in A ¯-chains. Consider intermediate movement to a noncriterial position. How is movement triggered to, say, an intermediate C?
(3)
One possibility is to allow non-feature-driven movement: movement is in fact free, but chains thus created must meet the requirement that the final site is marked with a criterial feature. An alternative more in line with current ideas, and argued for in detail in McCloskey 2002, is to assume that interpretable criterial features have purely formal counterparts (such as the ‘‘pseudo’’ wh-feature discussed by McCloskey) and that intermediate movement is triggered by such purely formal features, which bring the moved phrase closer to the target, in accordance with locality principles. Capitalizing the labels for substantive criterial features (Q, Top, Foc, R, etc.) and lowercasing the labels for the corresponding purely formal features (q, top, foc, r, etc.), we have the following representation for a sentence like (3):
(4)
We thus admit only one kind of movement, feature-driven movement (cutting across the distinction between substantive and purely formal features). Evidence favoring this view over the alternative admitting inter mediate non-feature-driven movement is the fact that languages often manifest morphological reflexes on the heads hosting intermediate landing sites in their Specs, as shown in the work briefly reviewed in section 5.5 (see McCloskey 2002 for a detailed argument based on Celtic A ¯ constructions). Agreement phenomena like past participle agreement in Romance, complementizer agreement in Celtic, and the triggering of I to-C movement are phenomena that would be expected if some feature is specified on the relevant head.
Given this discussion, I will tentatively assume that intermediate movement is driven by features that are the purely formal counterpart of substantive criterial features. There is only one kind of movement, feature-triggered; the triggering features are sometimes substantive (criterial, interpretable), sometimes purely formal (uninterpretable). Natural language syntax includes the latter featural option in order to implement intermediate movement, so that a potentially distant criterial position can be reached in successive steps, each of which satisfies locality requirements.
Now let us take a quick look at other A ¯-constructions. Relatives also involve a (sometimes overt) relative head and morphologically marked operators (if one assumes the head-raising analysis of relativization, then it should work in a manner akin to topicalization). Topic and focus constructions involve special heads, sometimes overt (see (9b–c)); I will assume that the head of the topicalized or focalized phrase is freely assigned the corresponding feature as it enters syntax, much like a f feature (so that no problem arises for the Inclusiveness Condition; see Chomsky 2000). This marking makes the phrase suitable for attraction and normally has PF effects reflected by the special intonational contours of topics and foci. So, the same derivational mechanisms can be uniformly assumed for all major A ¯-constructions, by and large the unification achieved by ‘‘On Wh-Movement’’ (but I continue to assume, in the spirit of the cartographic approach to the C system (Rizzi 1997, 2000, 2004a,b), that distinct landing sites are involved in the left periphery).
