John MacFarlane and Andy Egan have recently argued that a number of examples, "thirdparty assessments," show that the contextualist view on "epistemic modals" held by Hacking, Teller and DeRose is incorrect. They argue that the examples support a relativistic semantics for epistemic modality. I argue that not every utterance of a modal sentence involving 'may' or 'possible that' expresses the epistemic reading, that the problematic examples are cases where the utterance is ambiguous between epistemic and circumstantial readings, and that it is the circumstantial reading which drives the problematic third-party assessments. The treatment presented here is similar to that of John Hawthorne and von Fintel and Gillies. A major component of the thesis is a careful study of the way in which contextual factors affect both the "flavor" of possibility expressed by an utterance and the ability of a speaker to defend their modal claim against problematic third-party assessments.