Qianru Xu^1,2 (), Chaoxiong Ye^1,2, Weizhen Xie³; ¹School of Education, Anyang Normal University, China, ²Department of Psychology, University of Jyvaskyla, Finland, ³Department of Psychology, University of Maryland, USA

Visual working memory (VWM) is essential for temporarily storing and manipulating visual information, yet its limited capacity raises the question of how it can be used more efficiently. One potential mechanism is perceptual organization, such as grouping identical objects to help encode and retain more items. In this study, we used multivariate pattern analysis (MVPA) to see when and how perceptual redundancy affects the neural patterns underlying VWM. Participants performed a lateralized change-detection task with memory arrays containing all-same, partial-different, or all-different orientations. Time-resolved decoding revealed early neural differentiation (around 200 ms post-stimulus) between the all-same and other conditions, suggesting that grouping benefits arise already during the perceptual encoding phase, an effect that was not observed in our event-related potential (ERP) analysis. However, this effect was specific to the all-same condition, as partial redundancies yielded no intermediate neural benefit. Cross-temporal decoding further showed that although the all-same condition produced a stable and distinguishable neural pattern across the memory interval, the early neural differentiation did not generalize to later VWM stages, indicating a shift in the underlying representational code over time. Cross-condition decoding also revealed that the partial-different condition was represented more similarly to the all-different arrays than to the all-same ones. Together, these findings indicate that perceptual grouping in VWM operates through a separation-based mechanism of neural differentiation: full redundancy enables early compression into compact neural codes, whereas partial redundancy preserves separable representations to reduce interference. This separation-based compression account refines current theories of perceptual organization and highlights the value of MVPA for revealing representational dynamics that traditional ERP measures may miss. Future work should examine whether similar redundancy-driven mechanisms generalize to other visual features, such as color.

Acknowledgements: This work was supported by the Research Council of Finland (former Academy of Finland) Academy Research Fellow project (#355369 to Chaoxiong Ye) and Finnish Cultural Foundation (#231373 to Chaoxiong Ye, #240135 to Qianru Xu).