Abstract
Visual working memory representations interact with each other during maintenance. For instance, it has been repeatedly demonstrated that memory reports may be repulsed by (biased away from) a competing memory item nearby in feature space (Bae & Luck, 2017; Golomb 2015). Repulsion effects are thought to emerge because two similarly represented memory items may be more easily remembered via a mechanism that distinguishes the representations away from each other. However, a number of questions remain regarding the dynamics of these relational interactions. Here we asked: Are relational interactions rapidly computed during encoding and/or early maintenance and remain static, or do competing memory items continuously interact throughout maintenance until retrieval? In two experiments, participants studied the colors of two, simultaneously presented, real-world objects for one second. The colors of the two objects were randomly sampled either 45° or 90° apart in color space. After a delay of one second (Expt 1) or three seconds (Expt 2), participants were shown one memory item in grayscale at the center of the screen and asked to select the original color on a color wheel. Our reasoning was that, if relational interactions emerge due to continuous competition during maintenance, then a longer delay between study and test phases should increase repulsion effects. Alternatively, if similar repulsion effects are observed regardless of delay duration, this could suggest memory items are not continuously interacting during maintenance. Using probabilistic mixture modeling, we observed subtle yet robust repulsion effects in Expt 2, with color reports for the tested memory item biased away from the color of the untested memory item. Importantly, this effect was seen for a delay of three seconds but not for a delay of one second. These findings offer preliminary evidence that memory representations are sustained by dynamic interactions, continuously competing with each other during maintenance.
Acknowledgement: NIH R01-EY025648 (JG), Alfred P. Sloan (JG), NSF BCS-1632296 (AL)