Matthew Boyle (2011) has defended an account of doxastic self-knowledge he calls 'Reflectivism'. I distinguish two claims within Reflectivism, A) that believing that p and knowing oneself to believe that p are not two distinct cognitive states, but two aspects of the same cognitive state, and B) that this is because we are in some sense agents in relation to our beliefs. I find claim (A) compelling, but argue that its tenability depends on how we view the metaphysics of knowledge, something Boyle does not consider. I argue that in the context of the standard account of knowledge as a kind of true beliefwhat I call the Belief Account of knowledgethe claim faces serious problems, and that these simply disappear if we instead adopt an Ability Account of knowledge along the lines of that defended by John Hyman (1999, 2015). I find claim (B) less compelling, and a secondary aim of the paper is to suggest that once we reject the Belief Account of knowledge, and move over to an Ability Account, there is no explanatory role for (B) left to play.