EEG decoding reveals a link between visual working memory fidelity and the magnitude of similarity-induced memory bias
Poster Presentation: Tuesday, May 20, 2025, 2:45 – 6:45 pm, Pavilion
Session: Visual Memory: Neural mechanism of working memory
Schedule of Events | Search Abstracts | Symposia | Talk Sessions | Poster Sessions
Nursima Ünver1,2, Rosanne Rademaker3, Keisuke Fukuda1,4; 1University of Toronto, 2The Max Planck-University of Toronto Centre (MPUTC) for Neural Science and Technology, 3Ernst Strungmann Institute for Neuroscience, 4University of Toronto Mississauga
Visual working memory reports can be biased toward novel visual inputs presented during the delay. This bias is particularly strong when people report that these inputs look similar to the remembered target. In a prior study, we showed that the magnitude of this similarity-induced memory bias (SIMB) increased when certainty about the target decreased. This implies that the SIMB may be related to the strength of target encoding. To relate the magnitude of the behavioral SIMB effect to the neural fidelity of a target representation, we used EEG recordings together with multivariate decoding. Specifically, sixteen participants remembered the direction of a briefly (0.5s) presented dot motion stimulus (target) across a 2.5-second delay, after which they reported its direction. In one half of the trials, a second 0.5s dot motion stimulus (probe) appeared midway through the delay, and participants had to judge its similarity to the target. First, we replicated that when people judged the probe as "similar" to the target, target responses were biased toward the probe (i.e., the SIMB). Next, we successfully decoded the direction of target motion from the EEG for up to 0.4s after the target offset. We then hypothesized that, on trials in which the probe was judged to be similar to the target ("similar trials"), the magnitude of the SIMB would be large when the neural fidelity of the target representation is low. To test this, we median-split the similar trials based on the target decoding accuracy into the high- and low-fidelity trials and compared the SIMB magnitudes between them. Our results confirmed that the SIMB was reliably larger in the low-fidelity trials than in the high-fidelity trials. These findings demonstrate that neural fidelity, as reflected in decoding accuracy, is inversely related to the magnitude of SIMB.
Acknowledgements: This research was supported by the Natural Sciences and Engineering Research Council (5009170).