Evidence for a split between implicit and explicit attitudes is found most easily in cases where these attitudes diverge, and, mirroring a broader trend in dual process theory, researchers have tended to focus on cases where the implicit attitude is mistaken (or biased), and the explicit attitude correct. Philosophers and psychologists alike have attended more to cases of avowed egalitarians making biased decisions than to cases of agents with faulty overt principles but good instincts. This selective focus has made it harder to see what is distinctive about implicit attitudes. I argue that the line between normatively acceptable and unacceptable attitude formation is orthogonal to the line between implicit and explicit attitudes. I then examine the question of how that latter line should be drawn, given what we know about the ways in which these attitudes can be dissociated and manipulated.