We recorded from single neurons in visual area MT of the macaque in order to examine the spatial distribution of its directionally selective cells. The animals were paralyzed and anesthetized with nitrous oxide. All MT neurons (n = 614) responded better to moving stimuli than to stationary stimuli. For 55% of the neurons, responses to moving stimuli were independent of stimulus color, shape, length, or orientation. For the remaining cells, stimulus length affected the response magnitude and tuning bandwidth but not the preferred direction. MT neurons were divided into four categories on the basis of their sensitivity to moving stimuli: 60% responded exclusively to one direction of motion, 24% responded best to one direction with a weaker response in the opposite direction, 8% responded equally well to two opposite directions of motion, and 8% responded equally well to all directions of motion. The direction preferences of successively sampled cells on a penetration either changed by small increments or occasionally by approximately 180 degrees. Thus, there is a systematic representation of direction of motion. The representation of axis of motion, i.e., the orientation of the path along which a stimulus moves, is more continuous than the representation of direction of motion. There was a systematic relationship between penetration angle and rate of change of preferred axis of motion, indicating that cells with a similar axis of motion preference are arranged in vertical columns. Furthermore, axis of motion columns appear to exist in the form of continuous slabs in area MT. The size of these slabs is such that 180 degrees of axis of motion are represented in 400-500 micron of cortex. There was also a systematic relationship between penetration angle and frequency of 180 degrees reversals, indicating that cells with a similar direction of motion preference are also organized in vertical columns and cells with opposite direction preferences are located in adjacent columns within a single axis of motion column. Just as in macaque striate cortex where approximately 500 micron of cortex contain the mechanism for the local analysis of stimulus orientation, so in MT approximately 500 micron of cortex contain the mechanism for the local analysis of stimulus motion.