In the Stroop task word reading is thought to be automatic since it runs without intentional monitoring and is difficult to avoid. This view has recently been challenged by observations that Stroop interference is reduced when only part of the Stroop word is colored. In this study we asked whether the extent of Stroop interference varies with the position of the colored letter(s). We observed that Stroop interference was smallest when the first letter(s) were colored and largest when either the last letter(s) or whole word were colored. On these findings we suggest that colored and noncolored parts of partially colored words are processed separately and differently, and that selection of the color dimension for explicit report entails inhibition of the to-be-ignored colored letters.