The Freezing Effect of Agreement
Italian comparatives require a final glance. In Italian, while there is clear evidence that the wh-element quanto moves as a head (see (1)), the situation is complicated by some gaps in the paradigm: in particular, quanto never strands its complement (2a), except when it is cliticized with ne (2c) or when it shows no agreement with its adjectival complement (2b).
(1)
(2)
If we compare these data with those of Bulgarian, the generalization seems to be that agreement within the QP has a sort of a freezing effect, blocking head movement of the Q alone and the corresponding stranding of the remnant of the phrase. The same effect seems at work in the free relative cases discussed above: both in English and in Italian (3), the D head can never strand its nominal complement.
(3)
This kind of gap in the paradigm is very important for the hypothesis under discussion here, because it seems like a residue of a construction specific property of head movement, which would contradict our attempt at unification. We will see, however, that this constraint has nothing to do with head movement per se and that the same kind of effect occurs with phrasal movement. We will therefore end up confirming the claim that movement is a unitary operation. But let us proceed step by step.
Consider in (4) the (simplified) structure of a comparative construction containing an agreeing QP, like quante caramelle in (2a).
(4)
Suppose Q-N agreement means Q has an uninterpretable feature that needs to be checked by a feature of N. Head movement is not triggered here because it is useless: the head-head checking configuration is already there, merged in the base. Since agreement feeds pronunciation, this con figuration must be conserved until Spell-Out. This means that at the level of the first strong phase (vP), Q and NP must be adjacent. But at that point, VP, the sister of v, becomes inaccessible to the computation. There is therefore no way of moving Q without N: if it moves before Spell-Out, agreement cannot feed pronunciation; after that, it violates the PIC. The restriction on movement at play in (2) is due not to a peculiarity of head movement, but to the effect of a phonological adjacency requirement.
Crucially, a similar effect is attested with phrasal movement, con firming that it has nothing to do with head movement per se. As is well known, preposition stranding is prohibited in those languages, like French and Italian, where P and D tend to conflate in a PþD suppletive form (Law 1998; Salles 1997). In (5), this restriction against preposition stranding is shown at play in an Italian interrogative.
(5)
Consider the corresponding simplified structure in (6).
(6)
If P and D need to conflate, they need to be adjacent at Spell-Out. But at the first strong phase, when vP is sent to Spell-Out, VP becomes no longer accessible to the computation: the DP cosa cannot make a long movement alone, and preposition stranding is barred.
