Abstract
Introduction: There have been decades of work studying the computations and neural mechanisms by which cues predictive of the target location improve search performance. Often in these studies, observers are instructed to maintain fixation during search in order to isolate effects of covert visual attention from the influence of eye movements (Palmer et al., 2000). However, in the real world, human search involves the deployment of covert attention and also saccadic eye movements. Here, we measure the effects of cues on perceptual accuracy of humans during multiple-fixation search and compare them to the predictions of various foveated computational models (ideal searcher, Najemnik & Geisler, 2005; saccadic targeter, Beutter et al., 2003) which plan eye movements using the prior probabilities of target presence indicated by cues. Methods: Participants performed an 800ms search for a tilted Gabor among horizontal Gabor distractors embedded in spatio-temporal white noise. The target was located at one of sixteen locations, all equidistant from the center fixation cross (eccentricity = 9.3 deg). Either two, four, eight, or sixteen box cues indicated the possible locations of the target. In the experimental condition observers were allowed to move their eyes freely. In a control condition, observers were required to maintain central fixation throughout the duration of each trial. Results: Human improvement in search accuracy with fewer cues was greater for multiple-fixation search than for single-fixation search. Observer fixations clustered around cued locations suggesting strategic eye movement planning based on cue probabilities. The difference in cueing effects across single and multiple fixation searches were predicted by both the ideal searcher and saccadic targeting model. Conclusions: Humans use cues to strategize their eye movements which give rise to performance improvements beyond those of covert attention. This additional benefit from cues can be predicted from computational models of multiple-fixation search.