Abstract: What’s wrong with racist terms and other “defective” thick concepts? Answers to this question have been important in some recent discussions on conceptual role semantics (CRS) and the notion of epistemic analyticity. It has been argued that either CRS does not have the resources to explain what is wrong with pejorative terms (Williamson 2009), or that theorists must give up on the idea that meaning-constitutive inference patterns must be always truth-preserving (Boghossian 2003; Enoch & Schechter 2006; Eklund 2011). In this paper I will argue that CRS can deal with objectionable ‘thick’ concepts plausibly, and that the racist’s concepts are almost certainly not “blatantly defective” in the suggested way, at least given certain ways of filling out the CRS framework. This allows a defender of CRS and of the existence of epistemic analyticities one means of avoiding an important class of objections to these views.