Neural Systems Underlying the Implementation of Working Memory Removal Operations

Recently multi-voxel pattern analysis has verified the removal of information from working memory (WM) via three distinct operations replacement, suppression, or clearing compared to information being maintained (Kim et al., 2020). Univariate analyses and classifier importance maps indicate that some brain regions commonly contribute to these operations. This study aimed to use multivariate approaches to determine whether, within these commonly activated brain regions, each of these operations is being represented in a similar or distinct manner. To do so, we used Leiden community detection to identify brain networks that are characterized by similar multi-voxel patterns of activity with regard to these WM operations. Four networks were identified. The Visual Network shows similar multi-voxel patterns for maintain and replace, which are highly dissimilar from suppress and clear, suggesting this network differentiates whether an item is held in WM or not. The Somatomotor Network shows distinct multi-voxel patterns for clear relative to the other operations, suggesting that this network diff in clearing information from WM. The Default Mode Network has distinct patterns for suppress and clear, also suggesting that clearing information from WM is distinct from suppressing it. The Frontoparietal Control Network displays distinct multi-voxel patterns for each of the four operations, suggesting that this network has high involvement in regulating the flow of information within WM. These results indicate that how information is removed from WM depends on distinct brain networks that each have a particular manner in which their co-activation patterns represent these operations.