We have used the technique of PET to chart the areas of human cerebral cortex specifically responsive to an optical flow stimulus simulating forward motion in depth over a flat horizontal surface. The optical flow display contained about 2000 dots accelerating in radial directions away from the focus of expansion, which subjects fixated at the centre of the display monitor. Dots remained of constant size, but their density decreased from the horizon, lying across the middle of the screen, to the foreground at the lower screen margin; the top half of the display was void. For the control stimulus the dot motions were randomized, removing any sensation of motion in depth and diminishing the impression of a flat terrain. Comparison of the regional cerebral blood flow (rCBF) elicited by the optical flow and control stimuli was thus intended to reveal any area selectively responsive to the radial velocity field that is characteristic of optical flow in its simplest natural form. Six subjects were scanned, and analysed as a group. Four subjects were analysed as individuals, their PET data being co-registered with MRIs of the cerebrum to localize rCBF changes to individual gyri and sulci. There were three main areas of activation associated with optical flow: the dorsal cuneus (area V3) and the latero-posterior precuneus (or superior parietal lobe) in the right hemisphere, and the occipito-temporal ventral surface, in the region of the fusiform gyrus, in both hemispheres. There was no significant activation of V1/V2, nor of V5. These results show that higher stages of motion take place in both the 'dorsal' and 'ventral' visual pathways, as these are commonly conceived, and that both may be fed by area V3. The information potentially derivable from optical flow concerns the direction of heading, and the layout of the visual environment, a form of three-dimensional structure-from-motion. The perceptual division of labour between the various activated areas cannot be directly inferred, though it is a reasonable supposition that the parietal activation reflects the utility of optic flow for guiding self-motion.