To investigate the sources of spatial error in memory-guided saccades (MGS), we have trained monkeys on two different tasks: a MGS task and a delayed spatial match-to-sample (MTS) task. We first tested the effect of introducing a post-saccadic visual feedback on the accuracy of MGS. We found that visual feedback had a pronounced effect on the systematic saccade error, but less of an effect on the variable error. Visual feedback can improve the accuracy of saccadic eye movements over several days, while feedback removal can decrease accuracy in a reversible way. These effects also depend both on target eccentricity and the duration of the memory delay. To test whether saccade error is due to the accuracy of spatial memory storage or arises downstream from that memory, we measured behavioral performance on a spatial MTS task both before and after training with visual feedback. The results showed no significant difference in performance of the MTS task before and after feedback training despite significant changes in MGS accuracy. The results suggest that the accuracy of spatial memory is not the source of the systematic errors that accompany MGS.
