Displacement in Overt Syntax
overt displacement is the product of a two-step procedure:
(1)
a. Move F
The feature F (to be checked) of category a is extracted out of a and moves to the domain of a functional head H; F enters into a checking relation with an uninterpretable feature of H.
b. Pied-Pipe Category a is pied-piped to Spec,H.
Economy requires only Move F for the purpose of feature checking. If checking requires only the feature F (e.g., wh-or ø -features) of category a, then derivational economy should favor an operation (which mediates checking) that affects only F and not its entire category a. Thus, if there are no other considerations outside of feature checking, only Move F should apply, yielding no physically detectable effects of displacement. Pied-Pipe—which I argue follows the application of Move F—must be forced by some PF requirement(s); covert movement, for which PF considerations do not hold, therefore involves only Move F. The two-step procedure of Move F followed by Pied-Pipe is illustrated in (2).
(2)
Step (a) of (2) represents the application of Move F. H with an uninterpretable feature induces the matching feature F to raise out of category a once H is merged. I assume that when F is checked against the uninterpretable feature of H, the uninterpretable feature of H undergoes erasure, whereas F does not. F is generally interpretable at LF for the category out of which it was extracted (e.g., f-features of an NP, such as person or number features, or an interrogative/wh-feature of a wh-word, which functions for clause typing and yields the scope of the wh-word), so F is not subject to erasure since it yields certain semantic effects. Note that a ‘‘category’’ is understood as a set containing phonological, semantic, and formal feature matrices. I assume that lexical items enter the derivation from the lexicon with all of their features intact (as opposed to what hap pens under Distributed Morphology, where phonological features of lexical items are introduced later in the derivation (Halle and Marantz 1993)). I use the term category to refer both to a lexical item present in the structure and to a constituent that is composed of distinct lexical items. Note also that (2a) represents feature movement as targeting the root, whereby it creates a Spec of HP, as opposed to adjoining to the head H (cf. Chomsky 1995, 270). This is conceptually desirable, as it allows for all movement operations to be analyzed as cyclic, root targeting operations (see Fukui and Takano 1998 for arguments along the same lines).
Though sufficient for checking, the separation of F and its category in overt syntax gives rise to problems at PF. Properties of the phonological component therefore require F’s category to pied-pipe as well. Chomsky (1995) characterizes the situation as follows:
For the most part—perhaps completely—it is properties of the phonological com ponent that require such pied-piping. Isolated features and other scattered parts of words may not be subject to its rules, in which case the derivation is canceled; or the derivation might proceed to PF with elements that are ‘‘unpronounceable,’’ violating FI [Full Interpretation]. (1995, 262–263)
Under this approach, Chomsky suggests that overt movement is com posed of a complex of chain-forming operations:
Applied to the feature F, the operation Move thus creates at least one and perhaps two ‘‘derivative chains’’ alongside the chain CHF = (F;tF) constructed by the operation itself. One is CHFF = (FF[F];tFF[F]), consisting of the set of formal features FF[F] and its trace; the other is CHCAT = (a;ta), a a category carried along by generalized pied-piping and including at least the lexical item LI containing F. CHFF is always constructed, CHCAT only when required for convergence.... CHCAT should be completely dispensable, were it not for the need to accommodate to the sensorimotor apparatus. (1995, 265)
Thus, if F is extracted in overt syntax, its category must pied-pipe; otherwise, the resulting structure will be defective for PF. Given the statement quoted immediately above, there appears to be an assumption that in overt syntax the formal features remain as part of the category (Cat) that pied-pipes. However, there is another possibility: that F and Cat move separately (Move F first, followed by pied-piping of Cat) to the domain of H. I refer to the operation that raises Cat as Pied-Pipe (here I do not discuss the role—if any—of CHFF). Pied-Pipe raises Cat to a position where it is in a local relation with its formal features in the domain of head H. I propose that the relation holding between F and Cat is PF adjacency, a local relation required at the PF interface. PF adjacency is defined simply in (3).
(3)
‘‘Intervention’’ in this case is defined in terms of linear order: if X precedes both Y and Z, and Z precedes Y, then Z intervenes between X and Y.
Pied-Pipe applied to Cat therefore forms a Spec in HP (it targets the root), making F and Cat adjacent. Both Move F and Pied-Pipe apply cyclically and in a local fashion: F is moved only if there is an uninterpretable formal feature to be checked at the root category; Cat raises to the root category only if its feature F has been extracted, and F and Cat are nonadjacent. Both operations, Move F and Pied-Pipe, target the root category, creating multiple Specs. This was shown in (2), repeated here.
This approach to movement potentially allows the featural content of a category to be in two places at once, which is what we need to resolve the VMH paradox.
Let us redefine the VMH as follows:
(4)
Following the application of Move F prior to Spell-Out, we have the con figuration in (5). In (5), C is phonetically null and the arrow represents only the extraction of the wh-feature.
(5)
In (5), Pied-Pipe is blocked by a simple economy condition: Avoid superfluous steps. Who and its wh-feature are adjacent in (5), since no phonological material intervenes between them. Thus, adjacency between the wh-feature and who holds without the application of Pied-Pipe, so the operation is blocked in this environment.
Note that an EPP feature is unnecessary under this approach. PF considerations replace it as the driving force for overt category displacement.
