The effectiveness of cognitive strategies and the training procedure for a motor learning, which seemed difficult to control voluntarily, was examined. The task was the abduction without fluctuation of the big toe. Two experiments were conducted with 16 and 20 subjects in each. Results suggest that (1) it is necessary to provide a relatively long duration for each trial in the training session in order to let the subjects to find their appropriate strategies; and (2) the greatest facilitatory effect is obtained when they individually attempt to control the abduction of the big toe with attention to the direction or its movements which inevitably contract the M. abductor hallucis, whereas the knowledge of the location of the muscle has little effect on its motor learning. The implication of these results to the rehabilitation for physically handicapped persons was discussed.