Plausibly perceptual knowledge satisfies the following: (1) It is knowledge about things from the way they appear. (2) It can embrace more than the way things appear. (3) It is phenomenologically immediate and thus, in one sense, non-inferential. (2) and (3) place a significant constraint on adequate elucidations of (1). Knowledge about an object, from the way it looks, which embraces more than the way it looks, should not turn out to be inferential in the relevant sense. The paper shows how this constraint can be met, drawing upon a conception of a discriminative capacity. The discussion touches on literature dealing with observation in science, knowledge of other minds, and the possibility of moral knowledge.