Patrick Wu¹, Howard Egeth¹, Jonathan Flombaum¹; ¹Johns Hopkins University

Consider basic variations of visual search: (1) pop-out, where a target is defined by a unique feature, (2) directed conjunction search, where a target defined by a combination of a color and a shape is shown prior to the trial, and (3) undirected conjunction search, where the displays are identical to a directed search, but no information is given before the trial. These likely demand different computations to perform. But it can be difficult to identify how and why by looking only at reaction times and error rates. We conducted six experiments, modifying typical search by limiting stimulus exposure. By requiring participants to click as close as possible on a blank screen to the position of the target we could compute pixel distance from click to true location as a function of exposure. This approach produces curves that describe the microgenesis of a trial: the rate at which knowledge accumulates. Two key findings emerged. First, directed search curves are indistinguishable from pop-out, suggesting that knowledge of the target allows for suppression of visual features and detection of salient signals despite limited exposure. Undirected search curves were distinct from the others and did not show similar properties; knowledge accumulation (distance error as a function of stimulus exposure duration) in undirected search fit a parabolic model as compared to a logarithmic model in the others. Second, increasing the number of nontargets greatly impaired undirected search with hardly a measurable effect on directed search. After increasing the nontargets from 7 to 11, undirected search performance did not reach an asymptote even at 3 seconds of stimulus exposure. Undirected search appears to rely on processes that are different in kind from those that support pop-out and directed search.

Acknowledgements: Johns Hopkins University Provost's Undergraduate Research Award