Saccadic gain controlled by a visual discrimination task

Abstract

Saccade adaptation is a form of motor learning that maintains saccade accuracy in response to new sensorimotor contingencies. We know that reinforcement learning can induce saccade adaptation in the absence of a visual position error suggesting that conventional saccade adaptation might involve general learning mechanisms rather than only specific motor calibration mechanisms. Previously, an arbitrary reinforcer such as an auditory tone or stabilizing the target on the fovea has been successfully used to control changes in saccade gains. We now ask whether adaptation-like modulations in saccade amplitude may be induced by the ability to perform a visual discrimination task using a new gaze-contingent paradigm. An experimental paradigm was designed in which the post-saccadic retinal error was canceled by extinguishing the target as soon as the saccade onset was detected following the target step from the fixation location (45-degree upward and leftward with random amplitude). The visual background was entirely covered by '8' symbols of 7 by 7 pixels. At saccade offset, all background symbols transiently (60 ms) changed and participants were instructed to report which symbol was displayed (either 2,3,5 or E) in a 4AFC task. However the possibility to perform the discrimination task depended on the saccade horizontal gain: if a gain criterion (based on a moving median computed over the 50 previous trials) was not met one of four irrelevant symbols was displayed such that no correct response was possible. In two participants the criterion encouraged an increase in horizontal gain. The percentage of gain change computed with respect to the mean of the 200 first baseline trials revealed a mean horizontal gain increase of 53% for one participant and 21% for the other participant. We conclude that saccades are operant behavior that may be reinforced by the ability to perform visual discrimination tasks.

Meeting abstract presented at VSS 2017