In two experiments, participants named the color of a colored word, which was a Hebrew color word or a word in Hebrew that was different from a color word in one letter only. The magnitude of the Stroop effect increased with the location of the changed letter. It was smallest when the first letter of the color word was replaced, resulting in a noncolor word, and it was largest when the last letter was replaced. These results challenge the assumption that automatic reading, as indicated by the Stroop effect, can be explained exclusively by memory retrieval accounts of automaticity. The results also have implications for the sources of facilitation and inhibition in the Stroop effect.