Title: On the limits of elliptical repair: sluicing, clefts and LF-copy Abstract: This talk focuses on the precise form of the unpronounced syntactic structure in sluicing. In particular, I start out from the question of whether the structure underlying the sluice in (1) is a regular wh-question as in (2) or a cleft as in (3). (1) John loves someone, but I don't know who. (2) John loves someone, but I don't know who John loves. (3) John loves someone, but I don't know who it is. I show that Merchant (2001:115-127) convincingly rules out the possibility that *all* sluicing is derived from an underlying cleft, but that his arguments leave room for another scenario: sluicing derives from an underlying wh-question, but when such a question is unavailable, a cleft can be used as last resort. This scenario ties in nicely with a lot of recent work arguing that clefts can be used to salvage P-stranding violations under sluicing. Interestingly, however, this 'cleft rescue strategy' is not available in languages with morphological case marking on their wh-phrases. In this talk, I try to capture these data through a revised version of Chung, Ladusaw & McCloskey's (1995, 2006) LF-copy analysis of sluicing.