In a series of recent papers, Yalcin considers three alleged forms of challenge data for any descriptivist view about epistemic modals, so-called ‘epistemic contradictions’, de re epistemic modals, and our conflicting judgments in eavesdropper cases.  In undermining descriptivism, this data is to provide the central linguistic motivation for his own expressivism about modals. Here I show, first, that Yalcin has mischaracterized that data and, second, that, properly characterized, it poses no difficulty for the canonical, descriptivist view.  I then turn my attention to the question of why Yalcin’s initial presentation has struck so many as persuasive.  The source of the difficulty, I argue, is a poor understanding of when the truth-assessment methodology Yalcin and others have deployed is and isn’t sound.  Here I remedy that defect with a proposal for how we might test, for an arbitrary case, whether its deployment is sound.