In virtue of what is perception successful? In philosophy and psychology, we sometimes assume that visual accuracy amounts to a correspondence between percepts and subject-independent, physical properties. In this dissertation, I argue that we should reject this assumption in favor of norms grounded in the action-guiding nature of perception. Recent theories of perception purport to cast off the intellectualist baggage of twentieth-century thinking, and to address perception in its own distinctive terms. I show that these approaches are unified in aiming to reduce spatial aspects of the percept to subject-independent geometrical facts about the object-perceiver relation. In doing so, these views remain guilty of an unwarranted assimilation of perception to cognition. Perceptual constancy, the capacity to encounter a relatively stable world of object properties despite variation in sensory stimulation, is measured using a metric that has percept-physical property correspondence at one extreme, and retinal match at the other. Advocates of the correspondence norm freely redeploy this metric as gauging accuracy in perception, so that the closer a percept comes to invariantly matching the distal property, the closer it comes to veridically presenting the environment. Yet, correspondence views are committed to widespread misperception that cannot be accounted for in terms of evolutionary complexity. I distinguish between descriptive and normative enterprises in cognitive science, and suggest that we reinterpret the constancy metric as an empirically useful, descriptive quantificational tool—one that does not straightforwardly entail normative facts. With the correspondence norm undercut, I develop a more viable framework for understanding accuracy, one that draws on James Gibson’s ecological theory. Accordingly, accuracy is best understood pragmatically, in ecological terms such as usefulness. Partial constancy is often sufficient for an organism to act effectively in its environment, a result that suggests surprising consequences for what is seen in perception. In color ontology, there is some theoretical attention to descriptive facts about constancy. However, because of a worry about stipulating perceiver and context standards, theorists continue to reject ecological approaches to color. I resolve the worry by appealing to pluralism about scientific objects. The resulting framework is ecologically sensible, empirically useful, and deeply interdisciplinary.