There are in principle at least two binocular sources of information that could be used to determine the motion of an object towards or away from an observer: such motion produces changes in binocular disparities over time and also generates different image velocities in the two eyes. Existing psychophysical and physiological evidence is reviewed. It is concluded that these data are inconclusive concerning whether one or both of these sources of information are used in primate vision. Thresholds were measured for disparity modulations in dynamic (temporally uncorrelated) random dot stereograms (RDS), and for RDS in which the same random dot pattern was used throughout (temporally correlated). Although the first stimulus contains no consistent inter-ocular velocity differences, thresholds were generally slightly lower for this stimulus than for temporally correlated stimuli. Sensitivity to the temporal derivative of disparity is therefore adequate to account for human stereomotion detection. A stimulus was devised in which monocular motion was clearly visible to each eye (with opposite velocities) but in which all disparity changes were beyond the temporal resolution of stereopsis. This produced no sensation of motion-in-depth. Similarly, stimuli beyond the spatial resolution of stereopsis did not support stereomotion detection. These data strongly suggest that stereomotion is primarily detected by means of temporal changes in binocular disparity. We argue that there is no experimental evidence that supports the existence of a mechanism sensitive to inter-ocular velocity differences.