The Phase Impenetrability Condition and Successive Cyclicity
Before we turn to an account of the selective extractability of wh-phrases out of infinitive/subjunctive indirect questions, a few remarks are in order concerning the Phase Impenetrability Condition (PIC) and successive cyclic movement. Chomsky (2000, 2001, 2004) assumes the PIC (1) (adapted from Chomsky 2001, 14), which ensures that derivations proceed phase by phase, thereby reducing computational burden.
(1)
Chomsky claims that the PIC follows from the fact that Spell-Out is subject to the general condition on operations (2) (adapted from Chomsky 2001, 14).
(2)
In order to ensure successive-cyclic movement under the PIC, Chomsky makes the following assumption (adapted from Chomsky 2001, 12):
(3)
This provides an ‘‘escape hatch’’ for successive-cyclic movement through the edge of a phase.
Chomsky assumes the probe-goal theory of movement, where three kinds of uninterpretable formal features are involved in overt movement. In overt wh-movement to the Spec of an interrogative C, the following uninterpretable formal features are involved: the Q-feature of C, the wh feature of a wh-phrase, and the EPP feature of C, as shown in (4a).
(4)
The uninterpretable Q-feature of C, which counts as a probe, seeks a goal, namely, a matching feature. The Q-feature of C enters into a matching relation with that of the wh-phrase, which is interpretable. As shown in (49b), the Q-feature of the probe C and the wh-feature of the goal wh phrase, being uninterpretable, are both deleted (valued). The Q-feature of the wh-phrase, being interpretable, remains. As shown in (49c), the EPP feature of C, being a selectional feature, merges the wh-phrase in Spec,C. Since the EPP feature is uninterpretable, it is deleted (valued). This is essentially the mechanism responsible for overt wh-movement.
Chomsky (2000) applies the probe-goal theory of movement, which assumes three kinds of uninterpretable formal features to be involved in overt movement, to successive-cyclic movement without change. He assumes that a noninterrogative C or v without undeleted (unvalued) f features may also be assigned a nonspecific periphery feature (P-feature), which is contingent on the assignment of the EPP feature to the head of a phase in accordance with (3).
While essentially following Chomsky’s analysis of successive-cyclic movement, I depart from it in claiming that only C may be assigned an EPP feature (and a P-feature); v may not. Let us look at how successive cyclic movement proceeds under this analysis, taking (5) as an example.
(5) Where did John buy it t?
During its derivation, we construct the vP phase (6).
(6)
In (6), Agree of v with the object it establishes a Case-agreement relation, deleting (valuing) the f-features of v and the Case feature of it. Al though v is not assigned an EPP feature or a P-feature, where may raise to the vP-edge position, as shown in (7).
(7)
Although this movement is not triggered by any formal feature at this vP-phase level, it does not violate the economy condition that bans superfluous steps in a derivation. This is because according to (2), evaluation/ interpretation for the vP phase takes place at the next phase, namely, the CP phase. That evaluation includes whether or not movement of where to the vP-edge position satisfies the economy condition. At the CP-phase level, C is assigned an EPP feature and a Q-feature. These features trigger movement of where to Spec,C, which licenses movement of where to the vP-edge position. This ensures successive-cyclic movement without assuming that v may be assigned an EPP feature (and a P-feature).
The analysis of successive-cyclic movement proposed here should be preferred over Chomsky’s analysis on theoretical grounds, because it does not have to assume v with a P-feature. Chomsky (2000) assumes that P-features are defective, arguing that they are analogous to the [person] feature of a defective T. There is a case where C is assigned a Q-feature, a full complement of peripheral features. By analogy with the T-system, it is reasonable to claim that defective C with a P-feature also exists. In contrast to C, v is never assigned a Q-feature; this is clear from the fact that a wh-phrase can never be stranded in the vP-edge position. This casts serious doubt on Chomsky’s analysis, which assumes v with a P-feature.
