This study investigated effects of the target-background luminance ratios upon the autokinetic illusion, with special emphasis on manipulation of the background intensity. Exp. 1, in which the effects of four levels of target luminance were examined against the completely dark background, showed that the target luminance did not affect the illusion as long as the target was small enough in size (0.17 degrees in visual angle). This result confirmed the suggestion by Edwards in 1954. In Exp. II effects of the target-background luminance ratios were examined by varying the luminance of target and background independently. Dominant illusory patterns at the luminance ratio 1 were "pendulum-like" and "bobbing"; these differed from those at higher ratios ("winding"). On the other hand, latency and duration were not affected by the ratios. These findings suggest that the movement pattern is effective in specifying the autokinetic illusion, if it is appropriately categorized and represented.