MaxElide is a PF Constraint

Wh-movement out of the vP-ellipsis site is well-known to be blocked. Recent previous literature has attempted to capture this empirical fact with the MaxElide constraint. However, a new set of ellipsis data shows that the constraint overgenerates. We propose an alternative view of MaxElide, arguing that MaxElide applies at PF. The core issue is to explain why vP-ellipsis is impossible with intervening focus in root wh-questions, but possible in embedded wh-questions. We will first show that the previous analyses of MaxElide cannot resolve this contrast, and then argue that our alternative can. In both types of questions, the vP does not satisfy the semantic licensing condition, but only the TP does. The VP just looks like being deleted since the TP contains intervening focus. Following Merchant (2001), we suggest that the E(lliptical) feature cannot enter the derivation with C with T-to-C movement, which means [E] cannot enter C in root wh-questions. Consequently, neither the TP nor the vP can be elided in root wh-questions. This explains why ungrammaticality results in root wh-questions even with intervening focus.