Five human subjects viewed binocularly the interior of a full-field hemisphere filled with a random-dot pattern. During training, eye movements were recorded by the electrooculogram. A computer detected the end of every saccade and immediately moved the pattern horizontally either in the same or, in different experiments, the opposite direction as the saccade. The motion was exponential, its amplitude was 25% of the horizontal component of the antecedent saccade, and its time constant was either 25, 50, or 100 ms in different experiments. Before and after 2-3 h of this experience, movements of both eyes were measured simultaneously by the eye-coil/magnetic-field method while subjects made saccades across the moveable pattern, looked between stationary targets, or made saccades in the dark, to see the effect of such adaptation on postsaccadic eye movements. 2. After 2-3 h (10,000-20,000 saccades) subjects developed a zero-latency, postsaccadic, ocular drift in the dark in the direction of the pattern motion. Three subjects were trained to backward drift, two to onward drift. Drift amplitude in the dark changed by 6% of the saccade size (range: 2-11%). The drift was exponential with an overall time constant of 108 ms. 3. After training, while viewing the adapting pattern motion, the change in the amplitude of the zero-latency drift was approximately 10% (range: 6.5-14%). 4. Increasing the time constant of the pattern motion produced significant increases in the time constant of the ocular drift. 5. The incidence of dynamic overshoot (a tiny, backward saccade immediately following a main saccade) was idiosyncratic and went up in some subjects and down in others with adaptation. These changes did not seem related to modifications of postsaccadic drift. 6. Normal human saccades are characterized by essentially no postsaccadic drift in the abducting eye and a pronounced onward drift (approximately 4%) in the adducting eye. This adduction-adduction asymmetry is largely preserved through adaptation. Thus the changes in drift were conjugate and conformed to Hering's law of equal (change of) innervation. 7. These results agree with those previously demonstrated in the monkey and can similarly be explained by parametric changes in the pulse, slide, and step of normal saccadic innervation.
