This study investigated the information processing underlying adaptation to a changed sensory-motor relation. Twenty-two subjects performed a visual tracking task in which they were required to follow a target on a computer screen with a tracking cursor controlled via a joystick. The target stepped alternately up and down a constant vertical distance between two positions on the screen. Midway through the task, the gain of the relation between movement of the joystick and the resultant visual displacement of the cursor on the screen was altered without prior knowledge of the subjects, inducing a threefold increase in the sensitivity of the joystick control.

All subjects overshot the target by a large margin on the step following the gain change, but then rapidly came back on target without overshooting in the opposite direction as they corrected their error. This performance suggested a very rapid adaptation to the changed sensory-motor relation. Subject performance on the next target step was dependent on the time interval to its occurrence. If the next target step occurred within about one second, then the subjects’ initial response did not overshoot the target; but if the next step occurred after three seconds, then the subjects’ initial response again overshot the target.

This differential response dependent on the inter-step time interval demonstrated that there were two time courses of adaptation to the changed sensory-motor relation – a short-term mechanism and a long-term mechanism. The long-term mechanism appeared to account for learning the new sensory-motor relation and was manifested as a quasi-exponential decrease in the initial overshoot of the response across a number of target steps. The short-term mechanism functioned to maintain task performance and get the subject on target in spite of incomplete learning of the new sensory-motor relation. This mechanism appeared to cease operating once the response was settled on the target. A third level of adaptation also emerged, comprising increased reaction times and decreased initial submovement times, suggesting a change in movement strategy following the occurrence of the gain change. These adaptive mechanisms help to account for the robustness of human motor control in the face of constantly changing sensory-motor relations.