Abstract
Humans are able to learn statistical regularities from their visual environment to reduce interference from non-relevant, salient distractors. Previous studies have identified statistical regularities as spatial or feature-based, but the neural mechanisms of feature-based distractor suppression are still unclear. Also, the effect of distractor feature learning on visual working memory (VWM) performance has not been explored yet. To examine the impact of distractor feature learning on attentional selection and VWM performance, we implemented a variant of the additional-singleton task in which the distractor appeared more likely in one specific color (high-probability color) than other colors (low-probability colors). During this learning task, we simultaneously recorded the EEG. Before and after the learning task, participants performed a change detection task in which the high- and low-probability distractor colors were used. The behavioral results of the learning task showed a decrease in response times when the distractor appeared in a high-probability color compared to the low-probability colors, indicating learning of the distractor feature regularities. In line with the behavioral pattern, the neural measures revealed a larger target N2pc and a smaller distractor PD for the high-probability color, suggesting more efficient attentional selection when the distractor appeared in the more likely color. Interestingly, our data revealed a modification of the distractor PD over time: in high-probability trials, we found a decrease in the late PD while the early PD increased from the first to the second half of the experiment, indicating a temporal shift from reactive to more proactive distractor suppression. VWM performance, in contrast, was not affected by distractor feature learning. In summary, our results suggest that learned distractor feature regularities are proactively used to reduce distractor interference before the first shift of attention allocation without noticeably affecting VWM performance.