When QR Disobeys Superiority

1 Background

In recent minimalist theorizing, the operation of quantifier raising (QR) changing scopal properties of quantifiers is commonly viewed as having essentially a syntactic origin but a semantic effect. Bruening (2001) provides strong support for this view by investigating the effect of “scope freezing”: the scopal relations between two quantifiers one of which c-commands the other do not change when each of them undergoes QR. This is the case, in particular, in English double object constructions: the two quantifiers in (1a) do not exhibit scopal ambiguity. By a similar token, pair-list (P-L) answers expected when a quantifier moves over (a trace of) a wh-phrase (May 1985) are not elicited by wh-questions like (2a) (for those speakers who accept (2a) at all).

1. 1.

   1. a. The teacher gave a (#different) student every book.

      *∀ > a

   2. b. cf. The teacher gave a (different) book to every student.

      ∀ > a

2. 2.

   1. a. (#)Which student did you give every book? *P-L

   2. b. Which book did you give to every student? P-L OK

Elaborating on Fox’s (1995) idea that the level of vP is relevant for QR as the closest projection in which elements that have undergone QR may be interpreted, Bruening suggests that QR is subject to Superiority as derived from a version of the Attract/Shortest requirement on v. In (1), for instance, the (optionally selected) “P-feature” of v (in the sense of Chomsky 2001) attracts the higher quantifier first; it then attracts the second quantifier, which moves to an internal specifier in a “tucking-in” fashion (Richards 2001). The base-generated ordering of quantifiers is re-created upon QR.

• 3.

On these grounds, Bruening argues that QR operates according to a strictly syntactic mechanism (feature checking) and is therefore subject to regular syntactic constraints such as Attract/Shortest.

Presumably, then, we are dealing with the same constraint that rules out the Superiority-violating instances of multiple wh-movement.

• 4.

  1. a. Who bought what?

  2. b. *What did who buy?

Note that positing vP as QR’s domain of operation radically divorces QR from wh-movement, which operates at the CP level. Though in English both operations obey Superiority and so are both subject to Attract, the attracting heads in the two cases are different: v for QR, C for wh-movement. For Bruening’s proposal, this separation is necessary given the possibility of inverse scope in standard cases of quantifier interaction such as (5) where the inverse scope reading is captured via reconstructing the subject into vP together with raising the object by QR (see Fox 1995, 2000).

• 5.

  1. a. Some boy kissed every girl.

  2. b.

It is well known that in some languages (see, e.g., Rudin 1988), wh-movement appears to violate Superiority. This is suggested a priori by the pattern of forming multiple wh-questions in these languages, whereby all wh-phrases are fronted to the left periphery of the sentence in overt syntax, and, furthermore, the order of moved wh-phrases is free. The following example is from Russian:

• 6.

  1. a. Kto kogo videl?

     who who.acc saw

     ’Who saw whom?’

  2. b. Kogo kto videl?

     who. acc who saw

The existence of languages that apparently “violate” Superiority with multiple wh-movement raises the question of whether the same in principle could be true of QR, if the latter operates according to a strictly syntactic mechanism. In the context of the Superiority-”violating” languages, on Bruening’s account the fact that both multiple wh-movement and multiple QR show Superiority effects in a single language is simply a coincidence: all else being equal, if a language does not show Superiority effects with multiple wh-movement, we cannot predict whether it will show Superiority effects with multiple QR, and vice versa.

We argue that the distribution of multiple wh-movement and multiple QR is in fact more systematic than previously thought. In particular, the following two-way correlation holds:

• 7. A language shows Superiority effects with QR if and only if it shows Superiority effects with wh-movement.

This correlation reinforces Bruening’s view that QR is syntactic, as well as those...