Abstract
Previous research suggests that visual working memory (VWM) efficiency can be improved by grouping similar features together, thus allowing participants to ignore task-irrelevant features. This suggests that VWM processing can be feature-based. However, evidence from event-related potentials (ERPs) suggests that VWM processing can be object-based such that participants automatically process both task-relevant and task-irrelevant features. To better understand the circumstances under which feature-based processing occurs, we examined: 1) how task-irrelevant features are processed in VWM when there is a strong attentional set for task-relevant features, and 2) whether repetition among features within a display reduces the processing of task-irrelevant features. Participants completed two blocks of a change detection task. The strong attentional set was developed in the first block by only including color-change (task-relevant) and no-change trials. The second block included shape change trials (task-irrelevant), but participants were instructed to detect only color changes. In both blocks, some of the trials included feature repetitions in the pre-change arrays of four colored shapes; two different shapes shared the same color, and two different colors shared the same shape. Grouping similar features helped improve performance for task-relevant feature change trials but did not affect task-irrelevant feature change performance. Response time results revealed that task-irrelevant feature change trials produced significantly longer RTs, suggesting that the task-irrelevant feature impacted VWM processing but did not confirm which specific stage was impacted. ERPs, a more sensitive measure to confirm which stage of VWM processing is being impacted, suggest that the task-irrelevant feature was not processed during the comparison stage in VWM. Specifically, task-irrelevant change trials produced a minimal N200 amplitude compared to task-relevant change trials. Overall, the results from the current study suggest that processing in VWM, under certain circumstances, can be feature-based.