Neural Differentiation Underlying Perceptual Grouping Benefits in Visual Working Memory

Efficient use of visual working memory (VWM) is essential for storing and manipulating information to optimize moment-by-moment visual cognition under a capacity constraint. Yet, how such efficiency is achieved remains a vexing problem. For instance, when observers leverage perceptual organization to group multiple identical objects to retain more visual content, they can do so by grouping items proportional to the redundancy in perceptual inputs or by leveraging a separation-based strategy to maximize identical visual inputs from those that contain at least some discrepancy. Conventional evidence based on behavioral output or univariate event-related potentials (ERPs) could not distinguish these possibilities, motivating the current study to use a multivariate pattern analysis (MVPA) approach based on a published dataset to address this issue. In a VWM change detection task with scalp EEG recording, participants tried to remember visual arrays containing all-same, partial-different, or all-different orientation bars. Time-resolved decoding revealed early neural differentiation, around 200 ms after stimulus onset, between the all-same condition and the other conditions, suggesting that perceptual grouping benefits emerge early and exhibit a separation-based pattern. Cross-condition decoding further confirmed that visual arrays containing partially-different arrays were represented more similarly to the all-different arrays than to the all-same ones. Together, these findings suggest that perceptual grouping in visual working memory operates through a separation-based neural differentiation, in which full perceptual redundancy enables early compression into compact neural codes while partial perceptual redundancy preserves separable representations to reduce interference.