Neural mechanisms of feature binding in working memory

As a fundamental cognitive system with limited capacity, working memory (WM) strategically binds various features together to enhance its efficiency. However, the neural mechanisms governing feature binding in WM remain unsettled. Here, we employed functional magnetic resonance imaging combined with graph-based network analysis during a WM task in which participants maintained both color and location information throughout the delay period and subsequently detected and reported changes in color-location bindings versus individual features. Our results revealed a collaborative network that operates through a central workspace encompassing the somatomotor area, insula, and prefrontal cortex, underpinning the effective processing of bindings. Within these regions, we observed increased local efficiency and stronger connections during feature binding. Notably, connections within this workspace significantly correlated with behavioral performance. Among these regions, the somatomotor area, characterized by a shorter intrinsic timescale, responded more rapidly to visual input, carrying rich temporal information with more connections, and potentially served as the starting point during binding processes. These results highlight a dedicated workspace with sufficient and valid internal connections, facilitating successful binding through collaborative regional interactions.