Abstract
Purpose: Behavioral reverse correlation shows that the visual mechanisms driving perception and saccades utilize similar neural representations of the target. We suggested that diverging target representations would result in an inefficient coupling between eye movement planning and perceptual judgments (Eckstein et al., 2007). However, there is no theoretical analysis or empirical evidence to support the hypothesis that a mismatch in target representations would result in suboptimal search performance. Here, we use computational models of multiple fixation search (ideal searcher and ideal saccadic targeting) in conjunction with genetic algorithms (virtual evolution) to demonstrate the benefits on search accuracy of having similar neural representations of the target mediating saccade selection and perceptual decisions. Methods: We considered an ideal searcher (Najemnik & Geisler, 2005) and an ideal saccadic targeting model (Beutter et al., 2003) with different templates mediating the deployment of eye movements and the final perceptual decisions. Each template was a linear combination of Gabor functions representing V1 simple cells. We virtually evolved the saccade and perception templates to maximize accuracy in the final perceptual decision detecting a Gaussian target embedded in either: 1) white noise, 2) 1/f noise, and 3) 400 calibrated natural images. Results: Findings with white and 1/f noise confirmed that the genetic algorithm can converge to the optimal linear templates which for these statistically stationary Gaussian backgrounds can also be calculated from closed form expressions. Critically, the process of virtual evolution resulted in similar underlying templates for saccades and perception for all backgrounds and models considered. A mismatch between the saccadic and perception templates resulted in search performance deficits. Conclusion: The similar neural representations of the target in the human brain for saccades and perception optimizes search and might be expected to have evolved through natural selection in the neural systems responsible for visual search.