Perceptual evidence about the external world is paradigmatically defeasible. If something looks red to you, it is reasonable to believe that it is red, but if you are wearing rose-tinted glasses, it may not be reasonable at all to believe this, unless you have some independent source of evidence. In this paper, I will compare four models for how to understand this phenomenon. These models differ in their answers to two questions: what evidence we get about the external world through perception, and what our having that evidence consists in. I like one of these models better than the others, but in this paper my primary concern will be to compare their virtues and vices.