Abstract
Does saccadic gain adaptation, often described as a simple form of learning, have much in common with other forms of learning, in particular those guided by reinforcement? We investigated whether saccade adaptation could be guided by reinforcement if no retinal error existed. To eliminate retinal error, we either had the target vanish when the saccade started, or we placed the target on the fovea at the end of the saccade. To reinforce small saccades, we had subjects make saccades to targets 14 deg away, with an auditory signal after those saccades with amplitudes in the lower 35% of the previous 50 saccades. This auditory reinforcement caused a progressive decrease in saccade amplitude, much as in conventional saccade gain adaptation in which we displaced the target during saccades so that the fovea landed beyond the target. Alternatively, we used the reappearance of the target at the fovea after the saccade as the reinforcement, providing it only on the smaller saccades, as above. This too caused a progressive decrease in saccade amplitude. When we discontinued the auditory reinforcement of small saccades and instead reinforced those saccades with amplitudes in the highest 35% of the past 50 saccades, saccade gain returned to normal in most subjects, although in one subject the gain remained low, until we restored the retinal error by having the target reappear at its original location. We interpret the altering of saccade gain by differential reinforcement as suggesting that normal saccade adaptation may also be driven by reinforcement rather than only by average retinal error. More specifically, our finding that presenting the target on the fovea can act as a reinforcement suggests that normal saccade adaptation might be driven simply by the reinforcing value of those saccades that fall closer to the target.
This research was supported by a Fulbright Research Award to L. Madelain.