Figure 6.

Signal detection model for guided search. We assume that people have a priority map of the image in which regions are ranked from most likely to contain targets to least likely to contain targets (x-axis). Search involves sampling these regions in order from most to least likely. The distributions represent the likelihood that a region is actually a target or distractor. The area under the target curve is the number of image patches which contain targets, and the area under the distractor curve is the number of patches in the image without targets. The expected number of targets after searching a given number of patches (t) can be determined by finding the point on the x-axis where the sum of the cumulative distributions (shaded area) equals t (we call this point γ_t) and taking the cumulative distribution of the target curve up to that point (green shaded area).