Younger observers (M = 21 years) proved to be better than older observers (M = 68 years) at discriminating one direction of motion from another, highly similar one. Several days' practice steadily improved performance for both groups equally. Improvement was well restricted to the direction with which that observer practiced, and the full gains were retained for at least 1 month. Control measurements with various degrees of optical blur demonstrate that direction discrimination does not require a well-focussed retinal image. This rules out optical factors as the potential cause of the prepractice differences between groups.