Abstract
In numerous paradigms, from fear conditioning to motor adaptation, memory exhibits a remarkable property: acquisition of a novel behavior followed by its extinction results in spontaneous recovery of the original behavior. A current model suggests that spontaneous recovery occurs because learning is supported by two different adaptive processes: one fast (high error sensitivity, low retention), and the other slow (low error sensitivity, high retention). Here, we searched for signatures of these hypothesized processes in the commands that guided single movements. We examined human saccadic eye movements and observed that following experience of a visual error, there was an adaptive change in the motor commands of the subsequent saccade, partially correcting for the error. However, the error correcting commands were expressed only during the deceleration period. If the errors persisted, the acceleration period commands also changed. Adaptation of acceleration period commands exhibited poor sensitivity to error, but the learning was resistant to forgetting. In contrast, the deceleration period commands adapted with high sensitivity to error, and the learning suffered from poor retention. Thus, within a single saccade, a fast-like process influenced the deceleration period commands, whereas a slow-like process influenced the acceleration period commands. Following extinction training, with passage of time motor memory exhibited spontaneous recovery, as evidenced by return of saccade endpoints toward their initial adapted state. The temporal dynamics of spontaneous recovery suggested that a single saccade is controlled by two different adaptive controllers, one active during acceleration, and the other during deceleration.
Significance statement A feature of memory in many paradigms is the phenomenon of spontaneous recovery: learning followed by extinction inevitably leads to reversion toward the originally learned behavior. A theoretical model posits that spontaneous recovery is a feature of memory systems that learn with two independent learning processes, one fast, and the other slow. However, there have been no direct measures of these putative processes. Here, we found potential signatures of the two independent adaptive processes during control of a single saccade. The results suggest that distinct adaptive controllers contribute to the acceleration and deceleration phases of a saccade, and that each controller is supported by a fast and a slow learning process.
