Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz

Neural evidence of suppression failure in working memory under high semantic activation

Poster Session D - Monday, March 9, 2026, 8:00 – 10:00 am PDT, Fairview/Kitsilano Ballroom

Edward Leung¹ (), Jarrod Lewis-Peacock¹; ¹The University of Texas at Austin

Suppressing the maintenance of an item in working memory reduces immediate access to it and degrades the quality of its representation in long-term memory. Preliminary behavioral evidence has shown that the efficacy of suppression is hindered, however, for memoranda that promote semantic activation, e.g. by grouping items that share a semantic category. One possible mechanism could be that semantic activation changes an item’s representation into a format inaccessible to the suppression mechanism (e.g., chunking), resulting in failure of suppression regardless of the level of engagement of the cognitive operation. We address this hypothesis with multivariate pattern analysis of EEG data. On each trial, participants (N=34) first encode four images that were either grouped (high semantics) or interleaved (low semantics) by category, then are cued to either maintain or suppress one of the items, and finally respond to a memory probe. The key behavioral finding is that suppressed items on interleaved trials show poorer accuracy compared to maintained items (p=0.017, BF10=4.76) and compared to suppressed items on grouped trials (p=0.014, BF10=5.84). Furthermore, we demonstrate, for the first time in EEG data, successful decoding of maintenance suppression using power in six canonical frequency bands decomposed across 64 electrodes in 100 ms time bins. Interestingly, increased neural evidence for suppression engagement on grouped trials correlated with increased accuracy (failed suppression) compared to interleaved trials (p=0.067, BF10=1.34). These results provide new insights into the neural mechanisms by which semantic activation impairs suppression in working memory.

Topic Area: EXECUTIVE PROCESSES: Working memory