Ryan Williams^1,2 (), Susanne Ferber^1,2; ¹Carnegie Mellon University, ²University of Toronto

We examined whether the competitiveness of a given visual working memory representation, with regard to memory-driven capture (Study 1) and inter-item memory distortions (Study 2), is modulated by salience disparities present at encoding. In both studies, individuals encoded the colors of two slanted rectangles embedded within a broader cluster of vertically oriented rectangles. For the task-relevant stimuli, salience was operationalized according to the rotation of each item, with one (high-salience) item oriented to a 45-degree angle, which popped-out from the background, and one (low-salience) item oriented to a 12-degree angle, which was less differentiated from the background items. In Study 1, individuals had to either report the color of a probed item or perform a visual search task (unpredictably). To measure memory-driven capture, a task-irrelevant color-singleton was always present in the visual search displays, which could match the color of the high-salience item, the color of the low-salience item, or neither of the two memorized colors. Here, we found memory-driven capture to be limited to the high-salience item. To examine the role of encoding salience on inter-item memory distortions, in Study 2, the colors of the two memorized items were held to a constant difference of 45-degrees in a circular hue space. Additionally, probed items were tested using 2AFC judgments in which a visually similar lure was positioned either towards or away from the unprobed item in hue space. Through these judgements, we found representations of the low-salience items to be distorted by high-salience items (i.e., attractive bias), while low-salience items had no effect on representations of high-salience items. Overall, we demonstrate that, when salience-based differences are present at encoding, the most salient item holds a competitive advantage, in that, it more strongly influences which visual inputs are prioritized in the environment and skews the representations of less salient items.