This paper discusses the syntax of English variable negative concord (NC). Considering variation in acceptance and production of both object NC ('I didn't see nothing') and subject NC ('nobody couldn't see me'), we argue in favor of an amended implementation of the theory in Blanchette (2015), where the elements in NC dependencies are connected by movement. Addressing problems with the analysis of subject NC in particular, we propose that NC involves neg-raising of a null negative operator (cf. Zeijlstra 2004) which conditions the variable pronunciation of other heads in the structure with negative morphology.