In this paper I argue that because Churchland does not adequately address the distinction between high‐level cognitive theories and low‐level embodied theories, Churchland's claims for theory‐laden perception lose their epistemological significance. I propose that Churchland and others debating the theory‐ladenness of perception should distinguish carefully between two main ways in which perception is plastic: through modifying our high‐level theories directly and through modifying our low‐level theories using training experiences. This will require them to attend to two very different types of constraints on the modification of our perceptions.