Brad T. Stilwell¹, Darrell A. Worthy¹, Brian A. Anderson¹; ¹Texas A&M University

Salient items, such as uniquely colored stimuli, can be suppressed. Suppression is greater for a salient distractor that is frequently encountered. Such suppression has been argued to occur via statistical learning such that the priority of a distractor is down-regulated based on the predictiveness of its features. However, there are more generalized mechanisms that could lead to learned distractor suppression, such that each color becomes less distracting after each encounter via either feature-specific or feature-nonspecific decay. To test between these plausible mechanisms, participants searched for a unique shape target in the presence or absence of a salient color singleton distractor. Across all experiments, the frequency of the color of the singleton was manipulated to create one high- and several low-frequency singleton colors. Each experiment varied the manner in which these frequencies were realized: The high-frequency and each low-frequency color were presented either intermixed within blocks of trials or individually across blocks. When the colors varied across blocks, the ordering of blocks varied either unpredictably, in a fixed order, or were front-loaded so that all the high-frequency singleton blocks occurred before each low-frequency singleton block. We observed greater suppression for the high-frequency than the low-frequency colors across experiments. Fitting the RT data with computational models revealed that this learned distractor suppression was best explained by both a feature-specific and a feature-nonspecific exponential decay function; both models fit the data better than a model in which attentional priority reflected the frequency with which the color of the distractor was encountered, although some individual participants were best fit by this model. These results suggest that learned distractor suppression may be better explained as a skill that develops from experience suppressing stimuli that can be both feature-specific and feature-nonspecific, as opposed to a product of the learned predictiveness of distractor features.