Alexandre Fortuna¹ (alexandre.fortuna@unige.ch), Martin Constant¹, Dirk Kerzel¹; ¹University of Geneva

Attentional templates in visual working memory help guide selection during visual search, and retro-cues can change how much priority each stored item receives. This study used electroencephalography (EEG) to investigate how probabilistic retro-cues influence the dynamic interplay between visual working memory and attentional selection, hypothesizing that cue reliability induces graded, continuous adjustments in visual working memory prioritization rather than all-or-none shifts. Participants first memorized three colors. Following memorization, each color was associated with a retro-cue that signaled one of three probabilities (60%, 30%, or 10%) of that color being tested in the subsequent memory task. After a delay, participants completed a visual search task followed by a memory discrimination task. In the search task, participants located a target dot among two lateralized disks, where the distractor disk was sometimes colored to match one of the probabilistically cued VWM items, allowing measurement of attentional capture. The results did not reveal differences in N2pc amplitude across reliability conditions during visual search, and reaction times likewise did not vary as a function of cue reliability. However, memory discrimination performance improved gradually, with items linked to more reliable cues being judged more accurately and quickly. This demonstrates a graded distribution of memory priority rather than categorical differences Overall, these findings indicate that probabilistic retro-cues adjust the prioritization of items in visual working memory, as reflected in graded differences in memory discrimination performance. However, this prioritization did not translate into measurable differences in attentional capture or N2pc activity during the search phase, suggesting that probabilistic weighting influenced memory retrieval more strongly than online attentional selection.

Acknowledgements: This work is supported by the Swiss National Science Foundation (SNSF) with the grand number 219517