Social interactions are essential to our daily life. Here we test the hypothesis that social interactions, even a simple joint attention interaction (JA) in which one gets information from the other but does not need to respond, requires bidirectional communication. We further hypothesize that the two processes: feedforward (information from the 'sender' to the 'receiver') and feedback have different mechanisms. To test these hypotheses we apply a novel multivariate directed connectivity method on hyperscanning functional MRI data of human dyads during real-time JA interaction. The analysis enables obtaining directed functional connectivity within quadruplet networks at each time-frequency point along and after JA interaction. For a network of the right dorsal medial prefrontal cortex
(dmPFC) and the right temporoparietal junction (TPJ) of both 'sender' and ‘receiver’ brains, five pathways were observed at different times and of different structures. Three pathways corresponded to feedforward and two to feedback processes. Construction of multiple quadruplet networks resulted with the following findings: (1) Feedforward process favorited high fMRI frequencies while feedback favorited low frequencies. (2) There were more feedforward pathways. (3) There were more pathways which included the dmPFC compared with pathways that included the TPJ. (4) Regional involvements in the different processes (feedforward, feedback, low and high frequency) at the sending and receiving brains were different. These findings suggest distinct mechanisms for the inability to understand social signals (impairment in social interaction) and the inability to return signals (impairment in social communication) thus could help in diagnose and follow-up of social impairment subjects.