Frontal eye field activity before visual search errors reveals the integration of bottom-up and top-down salience

We investigated the saccade decision process by examining activity recorded in the frontal eye field (FEF) of monkeys performing 2 separate visual search experiments in which there were errors in saccade target choice. In the first experiment, the difficulty of a singleton search task was manipulated by varying the similarity between the target and distractors; errors were made more often when the distractors were similar to the target. On catch trials in which the target was absent the monkeys occasionally made false alarm errors by shifting gaze to one of the distractors. The second experiment was a popout color visual search task in which the target and distractor colors switched unpredictably across trials. Errors occurred most frequently on the first trial after the switch and less often on subsequent trials. In both experiments, FEF neurons selected the saccade goal on error trials, not the singleton target of the search array. Although saccades were made to the same stimulus locations, presaccadic activation and the magnitude of selection differed across trial conditions. The variation in presaccadic selective activity was accounted for by the variation in saccade probability across the stimulus-response conditions, but not by variations in saccade metrics. These results suggest that FEF serves as a saccade probability map derived from the combination of bottom-up and top-down influences. Peaks on this map represent the behavioral relevance of each item in the visual field rather than just reflecting saccade preparation. This map in FEF may correspond to the theoretical salience map of many models of attention and saccade target selection.