Human social behaviors involve complex interactions between individuals, and understanding how interbrain neural activity reflects and predicts these interactions is critical for advancing social cognitive neuroscience. While electroencephalography (EEG) hyperscanning has been widely used to explore interpersonal neural dynamics, most studies focus on pairwise regional coupling, overlooking the brain's intrinsic network-level organization. Here, we propose a spatiotemporal network analysis framework that combines Bayesian non-negative matrix factorization with EEG source imaging to identify interpretable subnetworks with spatiotemporal information. Applying this framework to dyadic EEG datasets from interactive decision-making tasks identifies eight task-relevant subnetworks, including the default mode network (DMN), somatosensory-motor network (SMN), and visual network (VN). Effective interpersonal coordination was associated with enhanced network-level time-domain interbrain synchrony and spatial-domain inter-subject similarity, and the fusion of these metrics reliably predicted interactive behaviors. Notably, synchrony and similarity involving DMN, VN, and SMN emerge as robust predictors of interactive behaviors, with spatiotemporal coupling most prominent within these subnetworks. These findings reveal spatiotemporal network signatures underlying interpersonal neural synchronization and demonstrate the importance of distributed subnetworks and their temporal and spatial alignment in achieving effective social interactions. This framework provides a useful computational tool for probing the neurobiological basis of social behaviors.
Li et al. (Fri,) studied this question.