In this paper I argue that syntactic categories are not part of UG but are acquired during first language acquisition as a result of doubling effects. In short, it is argued that a semantic operator OpF can only be (re-)analysed as a formal feature (that may yield a functional projection) if the language input exhibits doubling effects with respect to F. After providing a theoretical motivation for this hypothesis (the flexible formal feature hypothesis) I present a case study (negation and negative concord), which demonstrates that this hypothesis makes correct predictions about the syntax and semantics of (multiple) negations. I finally discuss some general consequences of the flexible formal feature hypothesis concerning clausal structure and the status of parameters.