Abstract
Background: Neural responses throughout visual cortex encode stimulus location in a retinotopic reference frame, and memory for stimulus position is most precise in retinal coordinates. Yet perception is spatiotopic: during natural vision, objects appear stationary despite eye movements. The purpose of this study was to test the accuracy of retinotopic and spatiotopic memory under different stimulus conditions. Method: We measured the accuracy of retinotopic and spatiotopic memory after saccadic eye movements under three different stimulus conditions. Observers fixated a dot and memorized the location of a brief peripheral memory cue. After a delay, the fixation dot changed position. Once observers made a saccade to the new fixation dot location, they were instructed to use a computer mouse to identify the location of the memory cue location. Observers were instructed to identify either the retinotopic or spatiotopic coordinate of the cue. In one condition, the display background was uniformly grey. In a second condition, a salient visual mask was flashed immediately after stimulus onset. In a third condition, a large background image of a natural scene appeared before the onset of the trial and remained on the screen until the end of the trial. Results: Retinotopic memory was more accurate than spatiotopic memory following an intervening saccade, replicating earlier observations. This accuracy advantage was eliminated when a salient visual mask was presented after cue onset, suggesting the retinotopic advantage was in part due to a retinal afterimage of the cue. Finally, we found that spatiotopic memory was more precise than retinotopic memory in the presence of images of natural scenes, suggesting that visual landmarks can be used by the visual system to maintain a representation of stimulus position across eye movements. Conclusions: Our results underlie the importance of visual landmarks in visual motor integration.