Abstract
Our internal representation of a complex visual scene relies on the dynamic processing of information in visual working memory (VWM). Though traditional methods have focused on storage limitations, here we move beyond these issues to explore cognitive abilities for dynamically manipulating information in VWM. Along the way, we discover independent limits for the storage and manipulation of visual items (Exp.1), spared memory for unmanipulated items (Exp.2), and preserved memory for the initial state of manipulated items (Exp.3). We developed a novel task in which participants were presented with a memory display consisting of colored circles, whose colors disappeared to leave behind circular placeholders. Pairs of placeholders swapped positions a varying number of times, after which participants judged the hidden color of a probed placeholder. This task is analogous to change-detection, along with a manipulation component. In Exp.1, we varied set size (2, 3, or 4 items) and number of swaps (0-4 swaps). We found a systematic impairment of memory when manipulating 3-4 items that increased with swaps. In contrast, performance with 2 items was unaffected by swaps (Fig.1). In Exp.2, we manipulated the number of times the to-be-probed item participated in swaps. Accuracy systematically decreased as the probed item participated in more swaps. In contrast, memory for unmanipulated items (0 participations) was better than memory for manipulated items – suggesting some protection for unmanipulated items (Fig.2). In Exp.3, we had participants retain the initial configuration during the manipulations, and we tested memory for both during the probe period. We found that memory for the initial configuration remained unaffected across all numbers of swaps (Fig.3). These results suggest two independent limits in VWM for storage and manipulation. In a dynamic world, the mind requires the ability to dynamically manipulate working memory representations – not merely the ability to passively store static representations.
Meeting abstract presented at VSS 2013