This dissertation offers a theory of what we know about our own actions when we act intentionally. The starting-point is an intriguing claim of Elizabeth Anscombe's: that necessarily, if an action is intentional, the agent will know that he is so acting, under the description(s) on which his behavior is intentional. Not only this, but he knows what he is actually doing without observation; he need not consult observational evidence about what he is bringing about. I call this the phenomenon of 'agential knowledge'. I connect this puzzling notion of agential knowledge to a contemporary debate about the nature of intention and practical reason. The major stances taken toward Anscombe's idea have tended to line up along opposing sides of this debate. On one side are J. David Velleman and Kieran Setiya, who agree with Anscombe that there is a deep action-knowledge connection. They argue that the best explanation of this connection is that intentions are a species of belief (a view I call 'strong cognitivism about intention'). On the other are those who hold that intentions are distinctively practical attitudes that do not constitutively involve the belief that you will do what you intend, or even entail that belief. The Distinctive Practical Attitude (DPA) theory of intention has been advocated by Donald Davidson, Michael Bratman, and Alfred Mele among many others. These theorists have been either skeptical toward the idea of a significant connection between acting intentionally and acting knowingly, or have simply not addressed the subject. I argue that neither of these stances is satisfactory: strong cognitivism about intention is implausible, and skepticism about agential knowledge is inadequate. As a solution, I propose an alternative theory of agential knowledge that is compatible with DPA views of intention. This account allows us to acknowledge that we do often have special, non-observational knowledge of our own intentional actions, without thereby being forced into a counterintuitive and problematic theory of intention, or to accept that agential knowledge is not supported by evidence. Starting with a suggestion of H. P. Grice's, which I criticize, modify, and expand upon, the central claim of my Inferential Theory is this: when an agent knows non-observationally that he is &phgr;-ing, that knowledge is evidentially based on his knowledge of his intention to &phgr;, together with his reasons for thinking his intention will be effective. The first task for the Inferential Theory is to secure the basis of the inference, by establishing that when an agent is acting intentionally, we can expect him to know what he intends to be doing. I draw critically on contemporary work in self-knowledge - most prominently, that of Richard Moran - in arguing that the conditions under which an agent's behavior counts as intentional are also conditions which generally afford him knowledge of what he intends to be doing. phi-ing intentionally and knowing that you intend to &phgr; are connected by the mental event of deciding to &phgr;, usually in light of practical deliberation. Next, I give a detailed account of the transition between knowledge of what you intend and knowledge of what you are doing, or will do. I discuss a variety of background conditions that bear on the justification of believing you are doing what you intend to be doing. Further, I explain how inferential knowledge can still count as non-observational in the relevant sense. A major reason is that the knowledge is evidentially based upon background knowledge of one's abilities and of one's own mind, rather than on observation of the action in question. But further, I argue...