To what extent is what we perceive influenced by what we know? Although a large literature purports to show effects of knowledge, expectations, and other cognitive states on various aspects of perception, strong counterarguments have been advanced that these demonstrations are confounded by nonperceptual factors. For example, although letters are easier to recognize in meaningful words than meaningless letter strings, skepticism remains that such effects of knowledge on visual recognition mean that knowledge literally helps people see. In Experiment 1, a perceptual matching task is used to show that meaningful words look sharper than meaningless letter strings. In Experiments 2 through 4, it is shown that people are more accurate in detecting subtle changes in blur when they occur in meaningful words compared with meaningless letter strings. In Experiment 5, it is shown that this improvement in performance cannot be explained solely by differences in visual familiarity, but is predicted by semantic factors such as word imageability. These results provide a strong empirical rejoinder to claims that perception is encapsulated from knowledge.