Optimal and non-optimal fixation selection in visual search W S Geisler1, J Najemnik2 |
|---|
1Center for Perceptual Systems, University of Texas, TX, United States |
| Under some circumstances humans are very efficient at fixation search. For example, in practiced observers, both the search time and the statistics of fixation locations and saccades have many characteristics of an ideal searcher, when the search target is known and the background is either a uniform field or a field of white or 1/f noise. However, there are circumstances where humans do not make optimal eye movements, and there are many more situations where one would not expect them to. I will discuss at least two cases. One case is at the start of the search trial where the observer may prepare for an eye movement, or eye movements, before the trial starts. The second case is search in complex displays, even with a single known target. Optimal fixation selection depends on the observer having an estimate of the detectability of the target at different locations across the visual field. When the background is complicated (highly non-stationary) the computational demands of estimating detectability can be very high. Under such circumstances humans are almost certain to adopt simpler (non-optimal) fixation selection strategies. [Najemnik and Geisler, 2005, Nature 434, 387-391; 2009, Vision Research, 49, 1286-1294; 2008, Journal of Vision, 8, 1-14] |
| Up Home |
|---|