Eva Lout¹, Jarrod Lewis-Peacock¹; ¹University of Texas at Austin

Visual working memory (VWM) supports the storage, updating, and use of goal-relevant information for adaptive behavior. Prior work indicates that updating a memory item may not fully overwrite its original representation, allowing both the old and updated representations to coexist. We had hypothesized that repeated updates may clutter VWM with old representations that degrade memory performance but instead found that increases in error are driven by the size rather than the number of updates. This suggests that memory degradation stems from imprecision in the updating process itself, rather than interference from older representations. Here, we asked whether incentivization could enhance the precision of updating. Rewards have been shown to improve VWM precision for stored items, but it remains unknown whether this benefit extends to transformed items. Participants completed a VWM task in which a remembered item needed to be mentally rotated a varying number of times (0, 1, or 2) and by different magnitudes (60°, 90°, or 120°). They earned reward points based on their memory precision on the first half (Experiment 1; N=17/30) or the second half (Experiment 2; N=16/30) of trials. We predicted that reward would enhance performance overall and may offset the costs associated with larger updates. Preliminary results show a statistical trend for improved performance with reward in Experiment 2, but not in Experiment 1. However, rewarded performance did not differ between experiments, suggesting the benefit of reward persisted even after it was removed in Experiment 1. Importantly, reward did not eliminate the magnitude-dependent cost of updating. These results suggest that reward can enhance the fidelity of transformation in VWM but cannot remove the costs imposed by larger changes. Thus, the mechanisms underlying VWM manipulation are precision-limited, and may underlie constraints on adaptive behavior in dynamic environments.