Effective connectivity directionality and hemispheric lateralization in working memory functioning in humans

Working memory (WM) in humans relies on dynamic interactions among distributed brain regions, particularly within the fronto-parietal network (FPN). However, most effective connectivity (EC) studies focus narrowly on canonical FPN nodes, overlooking contributions from other systems. Therefore, fundamental questions remain regarding directionality and hemispheric lateralization of WM-related EC, especially in healthy samples with sufficient statistical power. To address this, we applied Dynamic Causal Modeling (DCM) to fMRI data collected during a visual n-back task from 1091 healthy young adults in the Human Connectome Project (HCP). Regions of interest (ROIs) included bilateral FPN nodes (dorsolateral prefrontal cortex [DLPFC] and posterior parietal cortex [PPC]), the anterior insula (aIns) from the salience network, and the dorsal premotor cortex (PMd), given its role at the intersection of motor planning, cognitive control, and WM. EC during a control 0-back condition was excitatory between all ROI pairs, except for inhibitory influences originating from PMd. WM (2-back condition) strongly increased excitatory connectivity originating from PPC and aIns, and inhibitory connectivity originating from PMd. Notably, backward modulatory effects were largely absent for connections originating from DLPFC, apart from the left DLPFC→aIns connection. WM-related modulations were significantly stronger in the right hemisphere, indicating hemispheric dominance, while the overall pattern of EC was largely shared across hemispheres. These findings clarify the directional and hemispheric architecture of WM-related EC, confirming aIns and PMd as central integrative hubs. They support a distributed cross-network model of WM that extends beyond the traditional FPN framework.