Negative concord in fragments: Reexamining the evidence against the negativity of negation markers

I argue that particular restrictions on the recoverability of elided negation in fragment answers in negative-concord languages follow from the semantics of questions. Drawing parallels with other ellipsis phenomena, I also highlight the role of Agree in making certain heads recoverable. This invalidates a major argument against the analysis of negative-concord items as universal quantifiers scoping above negation.