The direction of motion in depth of a monocularly-viewed rigid sphere can be quantified in terms of the distance by which the sphere's centre will miss the centre of the pupil of the observing eye. If we express this distance as ns (where s is the sphere's radius and n is a scaling factor), then n approximates the ratio (d phi/dt)/(d theta/dt) between the translational velocity (d phi/dt) and the rate of expansion (d theta/dt) of the object's retinal image. To use this monocular information alone as a basis for motor action, prior knowledge of s would be necessary. (However, the value of s is available from binocular information, so that the distance by which the sphere's centre would miss the eye is, in principle, available from retinal image information alone and, in particular, without knowing the object's size or distance from the eye). We measured the just-discriminable difference in the direction of motion in depth for a monocularly-viewed simulated object. Thresholds were measured for trajectories contained within the horizontal, vertical and two oblique meridia. The translational speed of the retinal image was removed as a reliable cue to the direction of motion in depth by randomly varying the simulated object's speed on a trial to trial basis. The direction of translational motion was also removed as a reliable cue. Discrimination threshold for the stimulated direction of motion ranged from 0.03 to 0.12 deg for our seven subjects, and did not vary appreciably with the direction of motion relative to the line of sight over the range investigated, nor did it depend on whether trajectory was contained within the horizontal, vertical or oblique meridia. We conclude that subjects are able monocularly to discriminate differences in the direction of motion in depth, even when both the direction and speed of retinal image translation are removed as reliable cues.