Free Relatives and Comparatives
Good evidence that the phrasal status of wh-movement is only a byproduct of the environments where it occurs (interrogative CPs) would then be an instance of wh-movement embedded in a different environment, notably in a context compatible with DP selection and showing no phrasal pied-piping.
This is exactly what we find with free relatives. Consider the minimal pairs in (1a–b) and (2a–b) in Italian and English, respectively.
(1)
(2)
At first sight, free relatives (a,c) and interrogatives (b) differ in that the wh-element needs to be bare in the former but not in the latter. This anti-pied-piping restriction correlates with a different interpretation of the two constructions: as is well known, free relatives are complex nominal structures, while interrogatives are simple clauses. This is shown straight forwardly by the selection possibilities of the two structures: free relatives like (1a,c) or (2a,c) are compatible with a DP-selecting verb like mangi are (3a) or visit (4a); interrogatives like (1b) or (2b) are compatible only with non-DP-selecting verbs like chiedersi (3b) or wonder (4b).
(3)
(4)
This correlation can be expressed through a generalization like (5).
(5) A simple wh-structure excludes pied-piping exactly in those cases in which it occurs in a nominal position.
Given what we already know about heads and phrases, we can suggest a principled explanation for (5). What happens in free relatives is that the wh-element moves as a head. This means that, besides checking the wh feature on C as usual, it endows the clause with the D-feature required for interpretation.
Consider (6) for a more detailed comparison of free relatives (6a), full relatives (6b), and interrogatives (6c). The basic structural differences are minimal, but they correlate with important interpretive and syntactic differences. Relative clauses are characterized by the fact that they are complex nominal clauses (a DP embedding a CP: Kayne 1994; Bianchi 1999), the difference between free and full relatives being that D is moved in the former and merged in the latter. Interrogatives and free relatives involve the same uninterpretable feature [wh] on C, which is checked through phrasal movement in the former and through head movement in the latter. The crossed arrows in (6a) and (6c) correspond to instances of movement violating the CUC: a phrase projecting (6a);11 a head not projecting (6c).
(6)
A second piece of evidence that we are on the right track comes from a different instance of the generalization in (5), represented by comparative clauses. As we will see, comparative clauses are wh-constructions dis playing all the typical properties associated with wh-movement, together with clear complex nominal properties. We thus predict that comparatives should disallow phrasal pied-piping and involve wh-head movement. This prediction is confirmed, as the following review of the comparative construction’s properties will show.
First, comparatives are wh-constructions, displaying all the typical properties associated with movement since Chomsky 1977. For example, they are sensitive to island effects, as (7) and (8) show for Italian and English. In Italian, the comparative is moreover introduced by an overt wh-element.
(7)
(8)
Second, comparatives display clear complex nominal properties, as has been observed primarily on the semantic side. In particular, they give rise to the kind of scope ambiguities typically associated with nominals, as the following classic example from Russell 1905 shows:
(9)
The easiest and most standard way to account for the ambiguity of (9) is to derive it from the scope possibilities of the comparative clause, so as to include or exclude the belief verb from the comparison, as informally rep resented in (9a–b) (von Stechow 1984).
The complex nominal character of comparatives is also visible in their interpretation, which is that of amount or degree descriptions such as (free) relatives; in this way, they differ from corresponding wh interrogatives, which are simple amount or degree predicates (e.g., Heim 1985; Donati 2000). Again, this is represented informally in (10).
(10)
On the syntactic side, comparatives are strong islands for extraction, just like all complex nominal clauses, but unlike standard wh constructions.
(11)
Given that comparative clauses are wh-structures and they display these nominal properties, the prediction is clear: comparatives should in volve wh-head movement. This is exactly what the data in (11) and (12) show, from Romanian and Bulgarian, respectively.
(12)
(13)
a. *Ivan izpi povece vino ot-kolkoto bira Maria izpi [t]
Ivan drank more water of-how-much juice Maria drank
‘*Ivan drank more water than juice Maria drank.’
b. Ivan izpi povece vino ot-kolkoto Maria izpi [t] bira
Ivan drank more water of-how-much Maria drank juice
‘Ivan drank more water than Maria drank juice.’
c. Kolko bira izpi Maria [t]?
how-much juice drank Maria
‘How much juice did Maria drink?’
d. *Kolko izpi Maria [t] bira?
how-much drank Maria juice
‘*How much did Maria drink juice?’
(Izvorski 1995)
(12) and (13) show that the very same wh-element (cıˆt in Romanian, kolko in Bulgarian) moves as a head in comparatives (a–b), but as a phrase in interrogatives (c). This provides good evidence that the wh-element involved in the embedded comparative is a Q head, which endows the clause with the Q-feature required for interpretation. The same evidence holds also in English and Italian. In English, however, the facts are partially obscured since the wh-element is null and the contrast with interrogatives is not minimal.
(14)
In Italian, as we will see in the next section, the movement pattern illustrated so clearly by the Romanian and Bulgarian data is complicated by a gap in the paradigm that requires an explanation.
All the data discussed in this section provide strong evidence for the following analysis: Comparatives and free relatives, on the one hand, and interrogatives, on the other hand, display a minimal syntactic contrast that correlates with a semantic difference. While all involve wh movement, triggered by the same feature and constrained by the same restrictions, free relatives and comparatives move a bare head, which changes the CP configuration domain into a DP/QP category, turning the clause into a complex nominal; interrogatives move a phrase, which leaves the CP configuration domain unchanged as a simple (interrogative) clause. This is what we expect if movement is an operation merging just enough material for convergence, and if heads and phrases differ only in their projection ability.
