French Predicate Anaphora, Comparatives, and Ellipsis

Abstract

AbstractWhile French does not display the kind of VP-ellipsis found in languages like English, it effects verbal anaphora by means of two constructions that have received relatively little attention in the literature. Both of these combine the generic ‘action verb’faire‘do’ with what appears to be a pronoun. The first one, which I callle-faireanaphora (LFA), uses the cliticle, whereas the second one, which I callfaire-çaanaphora (FCA), brings into play the strong pronounça‘it/that’. Taking such pronouns to be substitutive pro-forms devoid of inner structure cannot, however, explain why LFA, but not FCA, can appear in the standard of a comparative, a property that distinguishes English VP-ellipsis from VP-pro-forms likedo it/that/so, which behave like FCA. On the other hand, LFA, unlike English VP-ellipsis, does not allowwh-extraction from a VP-internal position outside the realm of comparatives, which leads to a paradox left unresolved in the literature. In this article, I argue that LFA involves ellipsis of a wholevP phase. Specifically, I argue thatfairein LFA, but not in FCA, is a (semi)auxiliary verb that takes avP complement and bears an E-feature lexicalized by the cliticle. I show that this correctly predicts that A-movement out of the ellipsis site is possible and that A-bar movement of a degreewh-phrase out ofvP in comparatives is made possible by the fact that French licenses Quantification at a Distance, a grammatical operation that allows degree quantifiers to escapevP before the next phasal head (C) is merged and triggers ellipsis. Otherwh-quantifiers that do not benefit from the Quantification at a Distance option are then expected to not be able to escapevP prior to ellipsis, hence run-of-the-millwh-extraction out of VP is correctly predicted to be incompatible with LFA. I then turn to what appears to be a related anaphoric phenomenon, which consists of the predicational copulaêtre‘be’ and the cliticle, which I calll’êtreanaphora (LEA), and show that it too is an instance of ellipsis in that it allows the full range of extraction phenomena out of the missing constituent introduced by the copula. I assume that in those cases in which the copula takes a small clause, it participates in an asymmetric structure where it lexicalizes Pred, a head that mediates between subject and predicate. In those cases in which the copula introduces a passivized VP, on the other hand, I assume that it lexicalizesv, an assumption in line with theories that posit multiple instantiations of predicate mediators (so-called ‘flavors ofv’). In both cases, the copula in LEA heads a phase, bears an E-feature lexicalized byle, and is argued to license ellipsis of its phasal complement. Thus, elements are allowed to escape ellipsis by moving to the edge of the phase headed by the copula, and this correctly predicts that LEA permits a wider range of extractions out of the ellipsis site than LFA does.