Emotional and Linguistic Features Predict Prefrontal Functional Connectivity during Ongoing Dialogues: An fNIRS Investigation

Identifying the neural bases of language has been a central focus in neuro-science since the pioneering case studies by Broca and Wernicke. Contemporary research has moved beyond classical modular models to conceptualize language as supported by a distributed network of anatomically and functionally interconnected brain regions. Yet, few studies have explored how these networks operate during spontaneous, real-life speech, limiting our understanding of language in natural contexts. In this study, we collected data from 84 individuals engaged in live conversations. Participants prefrontal brain activity was recorded using functional near-infrared spectroscopy hyperscanning, and functional connectivity was quantified via wavelet transform coherence. Dialogues were manually transcribed, and computational methods were applied to extract emotional and semantic/syntactic features from the speech data. Using linear mixed-effects models, we found that emotional content significantly predicted prefrontal functional connectivity [Formula] Among all emotional predictors, expressed anger was the most robust: higher anger levels were associated with reduced connectivity between the left middle frontal gyrus and the right inferior frontal gyrus [Formula]. While other semantic and syntactic features did not predict overall connectivity, degree assortativity--an index of linguistic structure--was negatively associated with connectivity between the superior frontal gyri and the left inferior frontal gyrus [Formula]. These findings highlight how both affective and structural properties of speech modulate prefrontal connectivity during real-life interaction. More broadly, this work demonstrates the potential of integrating computational linguistics with social neuroscience to uncover the neural mechanisms of real-life social interactions.