Modeling Efficient Serial Visual Search

Humans perform visual search fairly efficiently, finding targets within only a few fixations. Data from eye-tracked participants was subjected to a fixation by fixation analysis to pinpoint why participants tended to make fewer fixations than would be expected by chance. The goal of this paper is to present a computational model that performs visual search as efficiently as humans. The model varied several components that may have aided visual search: memory, search strategy, and degree of parafoveal vision. Two dependent measures were used to evaluate the model: number of fixations to find the target and the distribution of saccade amplitudes. The best fitting model suggested that the biggest contribution to efficient search came from larger parafoveal vision. Search strategy, however, accounted for the distribution of saccade amplitudes.