This paper explores ways in which agentive, deontic, and epistemic concepts combine to yield ought statements, or "oughts," of different characters. I am especially interested in agentive ought statements whose violation invites criticism of the agent. I refer to these statements as "epistemic oughts," since an appeal to knowledge seems to play such an important role in their description. The investigation takes place in the setting of stit semantics, a modal framework for the analysis of agentive statements. I begin by supplementing stit semantics with an epistemic operator, and then exploring an initial account of epistemic oughts that results from combining this operator with agentive and deontic concepts in a straightforward way. After showing that this initial proposal is flawed, I then offer an account of epistemic oughts in which the role of knowledge is more complex, but which escapes the flaws of the initial proposal. Finally, I mention two directions for generalization: to relativistic oughts, and to conditional oughts. 1. This phrase has also been used to describe ought statements that are based on epistemic, rather than moral or practical, norms, such as "The keys ought to be in the office, since I've looked everywhere else," as well as ordinary ought statements that are themselves concerned with epistemic matters, such as "You ought to have known not to wait till the last minute." I do not discuss these other uses of the phrase here. 2. Histories of the subject, with references to the works of these writers and others, can be found in Segerberg (1992), and at various points throughout Belnap, Perloff, and Xu (2001).