Head-ing Toward PF

1 Introduction

In recent years Chomsky has put forward the idea that head movement is best viewed as a PF phenomenon. Consider, for instance, the following quotation (from a draft version of Chomsky 2000; see also Chomsky 1995:393n. 139, 1995:368, 1999:31, 2000:150n. 102).

Many questions dissolve if head-adjunction [head movement] is part of the phonological component. There are, I think, independent reasons to suspect that this may be true, at least over an interesting range . . .

Ever since the original minimalist paper (Chomsky 1993) and the adoption of the Extension Condition (Merge must always take place at the root of the tree), it has been noted that head movement patterns "differently"; and recent research ranging from purely theoretical (see Brody 2000, Freidin 1999) to psychological (Grodzinsky and Finkel 1998) has adduced evidence supporting that claim.¹ But note that "special character" does not in and of itself entail a PF treatment.² The only argument Chomsky offered up to 1998 (in class lectures) in favor of a PF treatment is that head movement does not seem to have any semantic import.³

In this squib we will provide an empirical argument in favor of the PF approach to head movement. Our discussion is based on pseudogapping constructions like (1).

(1) Debbie ate the chocolate, and Kazuko did eat the cookies.

In a series of papers, Lasnik (1995, 1999a,c,d) has appealed to pseudogapping examples to tease apart theoretical hypotheses pertaining to the controversial existence of overt object shift in English (movement of the object to [Spec, Agr_o]) and the competing definitions of strength. We will show that in the case of head movement pseudogapping again provides a special window on the workings of the computational system. In section 2 we review Lasnik's successive analyses of pseudogapping, emphasizing the role head movement plays in the explanation. In section 3 we point out several problems with the specific proposals Lasnik makes,⁴ and in section 4 we show how a PF treatment of head movement might help to solve them. While we want to stress that our aim is not to offer an explicit theory of how head movement might work in PF (aside from some remarks toward the end; see Boeckx, in progress, for further elaboration), we believe that the following discussion makes the PF analysis of head movement at the very least worth exploring.

2 Two Minimalist Approaches to Pseudogapping

Lasnik (1995) follows Jayaseelan (1990) in proposing that pseudogapping constructions like (1) result from VP-ellipsis, the remnant object having moved out of the VP, stranding the verb.⁵ However, Lasnik departs from Jayaseelan in taking movement of the object not to be a case of heavy NP shift, but one of overt raising to [Spec, Agr_o], as shown in (2) (setting details aside; see Lasnik 1995 for discussion).

(2) Debbie ate the chocolate, and Kazuko did [_AgroP the cookie_i[_VPeatt_i]].

The licensing of VP-ellipsis via overt movement of the object is Lasnik's (1995) crucial piece of evidence in favor of overt object shift in English. The question that immediately arises is why the verb need not raise in pseudogapping constructions, given that in nonelliptical sentences it must (assuming overt object raising; see the papers in Lasnik 1999b for extensive discussion and motivation).

(3) *Kazuko will the cookie_i eat t_i.

(vs. Kazuko will eat_i the cookie_jt_it_j)

This question is at the core of Lasnik's (1999c) discussion of the nature of strong features. Lasnik assumes that a strong feature is involved to force movement of the verb in (3). (2) and (3) seem to show that there are two possibilities for a convergent derivation. Either V can raise, presumably checking the relevant strong feature, or it can be part of the elided constituent. Lasnik notes that this mysterious state of affairs receives a straightforward account under the PF crash theory of strong features (Chomsky 1993).

This theory in essence says that an unchecked strong feature causes the derivation to crash at...