An agent’s degrees of belief should satisfy the axioms of probability. She should update her degrees of belief in the light of new evidence in line with the Bayesian rule of conditionalization. If she learns the objective chances, her degrees of belief ought to match them. In the absence of any evidence, she ought to distribute her degrees of belief equally over all possibilities. These are norms that govern epistemic agents when we represent them as having degrees of belief in the propositions they entertain. What establishes these norms? Pragmatic arguments have been given for some; evidentialist arguments for others. In this talk, Pettigrew wants to describe an alternative sort of argument. It begins with the claim that the sole fundamental virtue of degrees of belief is their accuracy, or proximity to the truth, and it provides a way of measuring this accuracy. Finally, it derives the consequences of this assumption. Amongst those consequences are the four norms just listed.