Xutao Zheng¹ (), Chenxiao Guan¹, Mowei Shen¹, Jifan Zhou¹; ¹Zhejiang University

Processing vast amounts of information with limited resources is a key challenge to human cognitive system. Visual working memory (VWM) is capable to retain only 3-4 objects. This number is even smaller in natural working memory tasks, where individuals need to repeatedly sample objects, simulating real-world scenarios. Research findings suggest that humans tend to minimize cognitive costs: when sampling cost is low, they sample more frequently but with fewer objects per sample to reduce VWM cost. In addition to minimizing costs, the voluntary utilization of VWM is supposed to maximize rewards for better cognitive outcomes in complex environments. This study adopts a copying task to examine how rewards influence the voluntary utilization of VWM. Participants copied six bars from a model to a workspace, autonomously deciding how many bars to copy each sample. The distance (long/short) between the model and the workspace determined sampling cost (high/low). In Exp.1, high or low rewards were assigned to the bars. We found high-reward bars were copied prior to low-reward ones and took longer time to copy, although accuracy reached a ceiling for both high and low-reward bars. Exp.2 replicated Exp.1’s findings, with the bars changed to the same color to prevent potential semantic encoding. Exp.3 introduced a time limit, then accuracy for high-reward bars was significantly higher than that for low-reward ones. Also, we found copying group for high-reward bars increased slightly as sampling costs increased, which should beneficial for maintaining accuracy. Exp.4 raised reward for copying multiple bars at once, leading to a significant increase in the VWM load despite high cognitive cost. In conclusion, rewards drive the adjustment of voluntary VWM operations to increase gains, even when it requires greater costs. Our research from a cost-benefit perspective could deepen the understanding of VWM's mechanisms and its real-world applications.

Acknowledgements: the National Natural Science Foundation of China (Grants 32371088, and 62337001), Science and Technology Innovation 2030—“Brain Science and Brain-like Research” Major Project (2022ZD0210800), and the Fundamental Research Funds for the Central Universities (226-2024-00118).