I argue that, if belief is subject to a norm of truth, then that norm is evaluative rather than prescriptive in character. No prescriptive norm of truth is both plausible as a norm that we are subject to, and also capable of explaining what the truth norm of belief is supposed to explain. Candidate prescriptive norms also have implausible consequences for the normative status of withholding belief. An evaluative norm fares better in all of these respects. I propose an evaluative (...) account according to which the goodness of true belief is, in Geach's sense, attributive rather than predicative. (shrink)