ALPHA POWER AND PHASE LOCKING AS MARKERS OF CONVERSATIONAL ENGAGEMENT IN TRIADIC INTERACTIONS

Oral communication is of central importance to human interactions. Speech comprehension required for oral communication has, however, so far mostly been studied in isolated, controlled laboratory conditions typically executed on single subjects. Here we sought to investigate naturalistic conversations between several participants that are more typical of real-world interactions. In particular, we aimed to identify whether EEG-derived neural markers can predict conversational engagement during naturalistic group conversations. In a hyperscanning setup, we recorded a multimodal dataset from 12 groups, each comprising three volunteers engaged in structured discussions that were elicited by three different types of conversation sparkers: consensus-building, conflict resolution, and free speech. Behavioral ratings of different aspects of conversational success were collected after each trial. EEG was acquired at 2,500 Hz alongside video recordings of participants' faces and their speech utterances captured at 44.1 kHz via directional microphones. All signals were temporally aligned and downsampled to 100 Hz prior to further analysis. From the EEG recordings, we extracted the spectral power in the alpha band and inter-brain phase-locking values (PLV) within predefined regions of interest during listening. We found that these measures could predict conversation success and showed differences between the three types of conversation. Our results demonstrate that hyperscanning-derived neural markers can offer objective indices of conversation quality and may, in the future, aid to assess the performance of hearing prosthetics in realistic environments.