Résumé : Social interactive learning is ubiquitous in human society. Learning through interactions with others plays an essential role in the daily lives of many people. Almost every day we communicate with peers or instructors, exchange information, observe and imitate other’s behaviors, with the purpose of learning. Yet how such interactive learning is parsed in the brains of interacting individuals remains poorly elucidated, and may be regarded as the “dark matter” of educational neuroscience. This thesis addresses the brain basis of interactive learning in three aspects: identification, modulation, and causation. Specifically, we investigated (i) whether interpersonal brain synchronization (IBS) could reliably identify interactive learning, (ii) how IBS and interactive learning could be modulated, and (iii) whether IBS plays a causal role in interactive learning. To approach these questions, we used functional near-infrared spectroscopy (fNIRS)-based hyperscanning to collect the neuroimaging data and naturalistic instructor-learner exchange as a principle experimental paradigm to realize interactive learning in the experiments. We developed a novel hyper-transcranial alternating current stimulation (tACS) system, which enabled manipulation of IBS between instructors and learners to examine the causative role of IBS in interactive learning. In the first study, we investigated the IBS between instructors and learners during more versus less interactive learning by computing the Wavelet Transform Coherence of instructor-learner brain activity. We have shown that IBS in the inferior frontal cortex reliably identifies and tracks interactive learning, and that IBS is more prominent when learning experience entails more turn-taking behaviors. Importantly, the unraveled IBS was associated with learning performance. In the second study, we further explore whether IBS and interactive learning could be modulated by the instructor’s verbal instruction. Results showed that compared to the explanation instruction, the scaffolding instruction elicited better learning performance and greater IBS. Besides, we adopted the machine learning to test IBS’s ability to discriminate between the two verbal instructions. We successfully showed that instructional strategies could be distinguished with a relatively high prediction performance. In the third study, we tested whether sleep deprivation (SD), which potentially impacts both social interactions and learning abilities, modulates interactive learning and IBS. Although learners performed below the baseline level immediately following SD, learning performance was comparable between sleep-rested (SR) and SD conditions after interactions with the instructor. Compared to SR, the SD condition induced greater IBS within instructor-learner dyads in the left inferior frontal cortex. Importantly, this IBS enhancement was associated with SD learners’ improved performance. Moreover, Granger Causality analysis showed that mean causalities from instructor to learner were significantly larger than vice versa following SD (but not SR). The fourth and final study investigated the causative role of IBS in interactive learning. By manipulating IBS through hyper-tACS, we demonstrated that externally induced IBS selectively biases interactive learning. Specifically, manipulation of IBS enhanced learners’ intonation performance and instructor-learner behavioral synchrony in a phase- (i.e., in-phase) and frequency-specific (i.e., 6 Hz) manner. The intonation performance also increased as a function of interpersonal synchrony. These results provide valuable insights into the functional role of IBS in interactive learning. The IBS may reflect the alignment of neural processes across learners and instructors. Such neural alignment impacts upon the acquisition of knowledge and information, and eventually upon the learning performance. Our studies hold strong relevance for real-world pedagogical practices and warrant future research to conduct clinical investigations that target learning deficits.