Recently, several theorists have proposed that we can perceive a range of high-level features, including natural kind features (e.g., being a lemur), artifactual features (e.g., being a mandolin), and the emotional features of others (e.g., being surprised). I clarify the claim that we perceive high-level features and suggest one overlooked reason this claim matters: it would dramatically expand the range of actions perception-based theories of action might explain. I then describe the influential phenomenal contrast method of arguing for high-level perception and discuss some of the objections that have been raised against this strategy. Finally, I describe two emerging defenses of high-level perception, one of which appeals to a certain class of perceptual deficits and one of which appeals to adaptation effects. I sketch a challenge for the latter approach. Suppose there is a cat lounging in a hammock. The weight of the cat's body pushes the fabric of the hammock down, creating a bulge in the bottom of the hammock. 1 In looking at this scene, what do you visually perceive? On one view, you perceive merely the shapes and colors of the cat and of the hammock. On another view, you visually perceive the colors and shapes of the cat and of the hammock, and you additionally perceive the cat as a cat, the hammock as a hammock, and the bulge in the hammock as caused by the weight of the cat. On this latter view, visual perception can represent a range of high-level features, since it can represent features other than the visual domain's paradigmatic low-level features of shape and color. 1 This example is from Siegel (2010, 121).