Vestibular function was examined during passive head movements having profiles that approximated low-to-intermediate range of natural self-generated movements (10-220 degrees/s peak velocity, about 0.5 s duration). A seated subject looked at a point target on the wall, the lights were extinguished and the chair was briefly turned while the subject tried to "look" at the just-viewed point. The chair was stopped, the lights were turned on again and the target was re-fixated, if necessary. Ocular stabilization was characterized (1) by "net stabilization" that was due to the combined effects of both slow-phase and rapid (saccadic or quick-phase) eye movements, (2) by "cumulative-slow-phase stabilization" that was due to slow-phase eye movements, and (3) by "cumulative-saccadic stabilization" that was due to effects of all rapid eye movements. It was found that both slow-phase and saccadic eye movements tended to keep the eyes on the actual unseen target. During repeatedly applied head movements, net and cumulative-slow-phase stabilization tended to be almost perfect. However, the average magnitude of the error in net stabilization (i.e., deviation from perfection) was always less than the corresponding error in slow-phase stabilization. This occurred because in a given turn, saccadic movements tended to supplement deficient slow-phase movements and to decrement excessive slow-phases. In 4 of 5 subjects, cumulative-saccadic stabilization tended to equal the error in cumulative-slow-phase stabilization. All results were unaffected by head velocities up to +/- 220 degrees/s. It was concluded that these saccades tended to stabilize gaze (eye + head) in space during head movements in total darkness.