Generative models of image features are a basic component of Bayesian theories of vision. How is generative knowledge used by the human visual system? I will describe behavioral and brain imaging results that are consistent with the idea that human visual perception of object properties involves the tight interplay between bottom-up and generative, top-down processes.