Abstract
We often have to search for multiple targets, and these searches often occur after substantial delays. These demands have implications for guidance from working memory (WM), and the brain mechanisms subserving search guidance. We conducted separate behavioral (eye movement) and fMRI investigations of the guidance process and mechanism. Our general method had subjects initially preview two objects (a butterfly and a house), presented sequentially. After 5000msec, this 2-object WM load was manipulated using 3 types of retro-cues (within-subjects variable): the letter “B” (only attend to the butterfly in WM), “H” (only attend to the house), or “E” (either the butterfly or house could be the target). Following a second 6000msec delay, a 9-object search display was presented and the subject's task was to indicate target presence/absence. Consistent with a WM load effect, errors were higher and RTs longer with an E cue compared to a B or H cue. Importantly, guidance was also affected by WM load, both in the number of distractors fixated before the target and in the proportion of initial target fixations. The identical experiment conducted in a 3 Tesla magnet revealed distinct posterior brain regions corresponding to the butterfly and house targets. Moreover, activation of these visual areas varied with the retro cue; one region showed higher sustained activation following the H cue, another region showed higher sustained activation following the B cue. Activation of both regions was observed following the E cue, but to a lesser degree. These fMRI analyses dovetail nicely with our behavioral evidence for guidance, suggesting that search guidance is mediated by WM representations in posterior visual areas. However, brain activation was also observed in prefrontal and parietal areas, suggesting the possibility of a larger network of systems contributing to the maintenance or reconstruction of the WM representations underlying search guidance.
This work was supported by NIH grant R01-MH63748 to GZ.