The arrowheads and lines of the Milller-Lyer illusion were varied independently in color and luminance. Stimuli were presented on a computer-controlled color television monitor. Subjects responded yes or no to a forced choice. Results indicated the following: (1) Differences in color between arrowheads and line reduce the illusion. (2) Luminance differences between arrowheads and line always reduce the illusion. (3) Effect 1 and 2 are independent of each other. (4) With a dark background, if the arrowheads are brighter than the line, the decrease is smaller than when the arrowheads are darker than the line. (5) A total change of luminance does not affect the illusion. Results 1 and 2 contradict Wickelgren's contrast hypothesis, which predicts a maximum illusion with the difference between arrowheads and background at maximum and the difference between line and background at minimum. Result 4, however, shows that an explanation in terms of similarity only is not possible. Therefore, a description in terms of both contrast and similarity is proposed.