Abstract
We investigated the relative contribution of bottom-up and top-down information to eye movements executed during a visual search. In two experiments, participants searched for a known target, a red (or blue) rotated T, among red and blue randomly rotated Ls. The stimuli were grouped into four 9-item clusters equidistant from a central initial fixation point. Bottom-up information was manipulated by varying the number of items in each cluster matching the target color (0, 2, 7, or 9 target-color items). Top-down information was manipulated by varying the likelihood that a particular type of cluster would have the target. In Experiment 1, the target was equally likely to appear in the 2, 7, or 9 (target-color) clusters, whereas in Experiment 2, every target-color element was equally likely to be the target. If search was simply guided bottom-up by the color composition of the clusters, then the search order though the cluster types would be the same in both experiments. Instead, initial fixations in Experiment 1 were directed to the 2, 7, and 9 clusters approximately equally, but in Experiment 2, clusters containing more target-color elements were more likely to be fixated sooner. In both experiments, the 0-cluster was rarely fixated and virtually never fixated after the initial fixation. Thus, other than avoiding clusters not containing the target color, the use of color was determined by the probability that the target would appear in a cluster of a certain color type. Once a cluster was fixated, however, the time spent within the cluster depended on the number of target-color items, consistent with a search of only those items. Thus, whereas within-cluster search was directly driven by color information in the cluster, between-cluster search was more indirectly influenced by color: only by color signaling the probability that the target was in a cluster.