Abstract
Previously we found that humans can approach optimal performance (in time and accuracy) when searching for a known target in backgrounds of broadband (1/f) spatial noise (Najemnik & Geisler, 2005, Nature, 434 387–391). However, there are a number of fixation selection strategies that can achieve similar levels of performance. These efficient strategies include fixating the location with the highest posterior probability of being the target (MAP searcher), fixating the location that maximizes the probability of correctly localizing the target after the eye movement is made (Ideal searcher), and fixating the location where the expected entropy of the posterior probability distribution is minimized (the ELM and EEM searchers). To determine which model most accurately describes the human fixation strategy, we measured the pattern of eye movements in a search task where a known Gabor target appeared in one of four possible pre-cued locations within a circular background of 1/f noise. All target locations were 4 deg from the initial fixation point at the center of the display. The two outermost locations were fixed at the right and left of the display, 180 degrees apart, while the middle two locations were randomly chosen on each trial from 9 possible equally-spaced locations between the two outermost locations. Following the cues, the subject initiated the display, which appeared for 500ms; this duration was picked to allow just one eye movement. The task was to indicate which of the four cued locations contained the target. We find that human observers fixate in a pattern that is highly dependent on the pattern of pre-cued locations and that they often make saccades to locations between potential target locations. The results are qualitatively consistent with the Ideal and EEM searchers, less consistent with the ELM searcher and completely inconsistent with the predictions the MAP searcher.