Abstract
What is the unit of visual working memory (VWM) representation? Feature-based account posits that features comprising an object are represented independently, but object-based account posits that those features are represented as an integrated format. We sought to tease these two accounts apart by examining whether distractors produce feature specific disruption. If the features are stored independently, distraction against a specific feature should disrupt that feature of representation, keeping the other features intact. However, if the features are integrated in VWM, distraction against a single feature should disrupt the entire representation. Participants (N = 26) were informed to remember target colors in one of two visual hemifields for a short period of time and adjust the color of a single cued item. We manipulated distractor features (color and location) combinatorically, resulting in four trial types. In one, targets and distractors shared the location set (location-repeated) or not (location-distinct), and in the other, they shared the color set (color-repeated) or not (color-distinct). Further, we varied the distractor context from baseline to distraction block. In the baseline block, distractors were absent to establish the natural unit of VWM. The distraction block followed where the distractor-absent and the four distractor-present trials were given. We analyzed response errors of the distractor-present trials with a mixture of precision and guess rate over the two phases of the distraction block. In the first phase, immediately after the baseline block, we found a feature specific disruption. The guess rate was lower, but the precision was worse in the color-repeated than the color-distinct trials. On the other hand, the guess rate was higher, but the precision was superior in the location-repeated than the location-distinct trials. However, the feature-specific disruption becomes nearly absent in the late phase. This adaptive behavior suggests that the distractor control facilitates integrating features of VWM.