Prey refuges are widely believed to prevent prey extinction and damp predator-prey oscillations. A review of the empirical evidence suggests that refuges are indeed capable of playing the former role. But the conditions under which they do so are not understood, nor is there any solid evidence for an effect on population fluctuations. The intuitive view that refuges act to stabilize equilibria and damp predator-prey oscillations is based in several theoretical studies of extremely simple models. Using a more realistic model, I show that several kinds of refuges can exert a locally destabilizing effect and create stable, large-amplitude oscillations which would damp out if no refuge was present. This finding contrasts sharply with the usual view. I argue that current evidence is tol weak, and the range of theoretically possible effects is too broad, to justify any simple characterization of refuge effects in nature. Manipulative empirical studies are an important first step toward correcting this situation, and I discuss some important factors to consider in their design.