Ellipsis and EPP Repair

1 Introduction

Merchant (2001) presents an interesting argument based on the lifting of Subject Condition effects under sluicing for the idea that ellipsis suspends the requirement that the subject raise to Spec,IP (the Extended Projection Principle, EPP). Lasnik and Park (2003) have taken issue with that argument and have called into question a conception of the EPP as a PF condition. After showing that Lasnik and Park's objections to Merchant 2001 are unfounded, we will present three novel arguments to the effect that the EPP is indeed suspended under ellipsis and that, on an approach to the ellipsis phenomena in question that takes them to involve PF deletion (or failure of transmission to PF, with concomitant nonapplication of Spell-Out; see footnote 3), the EPP is hence likely to be a PF condition.

2 The Sluicing Argument

2.1 The Argument

As is well known, extraction from subjects of finite clauses is impossible in English.¹ Thus, the sentences in (1) are ungrammatical. Those in (2), however, are perfectly well formed (as Chung, Ladusaw, and McCloskey (1995) first noted for similar examples). These are examples of sluicing, featuring ellipsis of the IP following the wh-phrase which (Marx brother). Apparently, sluicing eliminates the Subject Condition effect seen in (1). The question is how it manages to do so.

(1)

1. a. *Which Marx brotheri is [a biography of t_i] going to appear/be published this year?

2. b. *Which Marx brotheri did [a biography of t_i] cause a scandal earlier this year?

(2)

1. a. A biography of one of the Marx brothers is going to appear/be published this year—guess which (Marx brother).

2. b. A biography of one of the Marx brothers caused a scandal earlier this year, but Bill doesn't recall which (Marx brother).

Merchant's (2001) answer to this question runs as follows. Assume first of all that the Subject Condition is a constraint barring extraction out of the subject of a finite clause in its derived position, Spec,IP, and that there is no ban on extraction from the subject when it is in its VP-internal base position (whether it be the verb's complement position, as in the ergative/passive (a)-examples, or the specifier position,² as in the (b)-examples). Assume in addition that in sluicing contexts, the requirement that the subject raise to Spec,IP (the EPP) is suspended. Then the absence of Subject Condition effects in (2) is as expected: which (Marx brother) extracts from its containing noun phrase while the latter is inside VP (see (3)).

(3)

1. a. a biography of one of the Marx brothers is going to appear/be published this year—guess [_CP which (Marx brother)_i [_IPis [_VPgoing to be published [a biography of t_i] this year]]]

2. b. a biography of one of the Marx brothers caused a scandal earlier this year, but Bill doesn't recall [_CP which (Marx brother)_i [_IPI [_VP[a biography of t_i] caused a scandal earlier this year]]]

What is it that allows the subject to stay in VP in sluicing contexts, while this is evidently impossible in constructions in which no IP-ellipsis takes place? Merchant's answer is that the EPP is a PF condition. Specifically, the requirement that Spec,IP be filled is encoded in terms of an uninterpretable feature of I that is visible at PF and hence must be eliminated prior to PF, by movement to Spec,IP—if, that is, I itself remains present in the PF representation. If, on the other hand, I is elided at PF, as it is in the sluicing examples, then the EPP is not in effect, and the subject will perforce stay inside the VP.³ For extraction from the subject, this has the beneficial effect of legitimizing it in sluicing contexts.⁴

Lasnik and Park (2003) argue against Merchant's account of the contrast between (1) and (2) by showing that the idea that movement to Spec,IP does not take place in sluicing contexts causes trouble in connection with quantification...