On the syntactic flexibility of formal features

This paper argues that the set of formal features that can head a functional projection is not predetermined by UG, but is derived through L1 acquisition. I formulate a hypothesis that says that every functional category F is realised as a semantic feature [F] unless there are overt doubling effects in the L1 input with respect to F, in which case this feature is then analysed as a formal feature [i/uF]. In the first part of the paper, I provide a theoretical motivation for this hypothesis, and in the second, I test the proposal with a case study, namely the cross-linguistic distribution of Negative Concord (NC). I demonstrate that negation in NC languages must be analysed as a formal feature [i/uneg], whereas negation in Double Negation languages remains a semantic feature [neg] (always interpreted as a negative operator). This paves the way for an explanation of NC in terms of syntactic agreement. In the third part, I argue that the application of the hypothesis to the phenomenon of negation yields two predictions that can be tested empirically. First, I demonstrate how this hypothesis predicts that negative heads (Neg°) can be available only in NC languages; second, that independent change of the syntactic status of negative markers can trigger a change with respect to the occurrence of NC in a particular language. Both predictions are shown to be correct. Finally, I outline the consequences of the proposal presented here for the syntactic structure of the clause and also for the way in which parameters are associated with lexical items.