Head Movement versus Phrasal Movement
We are now looking for the property that triggers phrasal movement instead of head movement in the classical wh-constructions. The first place to look is in the definition of movement, among the conditions that constrain its various components. Let us review what these constraints on movement look like, given minimalist assumptions and the feature approach, to see if there are any that might be responsible for the choice between the two movements.
The traditional account relies on conditions concerning Agree, the search procedure—call them locality or closeness. Head movement, the default option, is limited by the Head Movement Constraint (HMC; Travis 1984). Whenever the goal is not (a feature of) the head of the complement of the probe, the nondefault option is chosen, and the goal pied pipes its phrase; the HMC has recently been reframed in these terms by Pesetsky and Torrego (2001, 363), as follows:
(1)
This is not a viable solution for both theoretical and empirical reasons. On the theoretical side, there is no room in a feature approach for such a constraint. Notice, interestingly, that Pesetsky and Torrego arrive at (1) by assuming that phrasal movement is the default option, blocked in case (a) because it results in remerging. But the system cannot work this way. If movement is triggered by a search procedure of a feature (Agree), locality must be defined in terms of features with no reference to the amount of material that gets moved with the feature. Such is the case with the Minimal Link Condition (2) and its more recent version (3).
(2)
(3)
In other words, locality is a condition on the search procedure, which is a component of movement defined at the feature level. At that level, the distinction between phrase and head plays no role and indeed does not exist at all. Therefore, the criterion we seek for choosing between head movement and phrasal movement cannot be found among the locality conditions.
On the empirical side, let us look at the well-known though controversial ‘‘long head movement’’ phenomena recorded in the literature (e.g., Lema and Rivero 1990; Manzini 1994; Roberts 1994; Carnie 1995; Bors ley, Rivero, and Stephens 1996). For example, the simple Breton sentence in (4) appears to instantiate the derivation in (5), where the verb raises to a left-periphery head, skipping the head I—hence violating the HMC, or the generalization (1).
(4)
(5)
The same derivation yields an ungrammatical result if the sentence contains a different kind of intervener, such as negation.
(6)
The explanation proposed by Roberts (1994) and others is that Relativized Minimality–type constraints such as (2) and (3) apply ‘‘across the board,’’ to any movement dependency, no matter whether it involves a head or a phrase. In other words, what counts for locality is the class of interveners and the class of attractees, not the amount of moved material. More data of this sort are given in (7) through (9).
(7)
(8)
(9)
The interpretation of such data in terms of long head movement is controversial and still being discussed, but the point is that these are exactly the data we predict given a feature approach to movement: if locality is a condition on the search procedure component of movement, then it is to be defined in terms of features, or classes of features. We thus predict that the same kind of constraints will govern both head movement and phrasal movement. But if locality is not responsible for the choice between head and phrasal movement, we must look at other movement constraints.
A second class of constraints on movement relates to the goal, which must be active (Chomsky 2000) and not embedded in an island. But these constraints do not seem to be relevant for present purposes, since they ap ply to any movement operation.
Finally, a third class of constraints is conceivable, relating to the other operation involved in Move, namely, Internal Merge. The Condition on Uniformity of Chains (CUC) can be defined as such a constraint.
(10)
(I will show in the next section that this constraint is only apparently irrel evant for the head/phrase distinction we are seeking.) Also conceivable is an economy condition on Internal Merge, for which what is to be merged is the minimal amount of material necessary for convergence: (11).
(11) Merge just enough material for convergence.
This is indeed the notion of minimality I referred to earlier, when I said that the minimal amount of pied-piped material is a head, giving head movement as the unmarked option.
The only conclusion we can draw from what we have seen so far is that there is no difference whatsoever between head movement and phrasal movement, except that the former is more minimal than the latter. To find a way to derive phrasal movement as an option and its selection in standard wh-constructions, we must turn to some other component of the grammar.
