 |  | |  | ||
                 | Visual search has no foresight Jeremy M Wolfe 1,2, Todd S Horowitz 1,2, David E Fencsik 1,2, Stephen J Flusberg 1 1 - Visual Attention Lab, Brigham & Women's Hospital, Boston, MA During an extended visual search task (e.g. search for T among Ls, "TvL"), does sensitivity to target identity accumulate gradually or are targets identified swiftly, but only once they are selected by attention? We used a novel, event-related technique that allowed us to measure signal strength at different times prior to the end of search. Os searched for horizontal Ts among rotated Ls and reported target orientation. Stimuli were scaled to be peripherally identifiable. Mouseclicks produced very brief glimpses of the display. Click positions served as confidence ratings scaled from “highly confident left” to “highly confident right”. A final mouseclick, localizing the target, ended the trial. Ratings were averaged as a function of number of frames prior to final response. Ratings were used to generate ROCs for each frame relative to response. Just two frames prior to finding the target, d' was near zero for most Os, even though they had searched for many frames. We compared these data to two control conditions: assessment of a gradual random walk toward a boundary (guaranteed slow accumulation of information); and TvL search with small stimuli that required foveation (guaranteed step function from no signal to perfect identification). TvL search data mimicked the mandatory eye movement condition, rather than the slow accumulation control. There was no evidence for gradual accumulation of information in this search task. Os have no explicit information about the location or identity of targets in TvL search task until attention selects the target. |