Abstract
Conversations-hallmarks of human social interaction-rely on a deeply rooted evolutionary tool: vocalization. Vocalization productions begin with an airflow from the lungs, shaped by the vocal folds and articulators of the mouth to generate a rich variety of sounds. Speech production therefore involves tight synchronisation between respiratory and articulatory events, especially in natural social interactions. During conversations, turn-taking indeed implies respiration anticipation: lungs must be filled with enough air to produce the next speech segment. Here, we use an fMRI corpus of live, unconstrained conversations to identify key brain regions involved in the respiratory anticipation of speech onset. Behaviourally, respiratory local maxima preceded speech onset ("Resp+" events) by approximately 200 ms, consistent with volitional control of breathing overriding automatic respiratory rhythms. Contrasting Resp+ events with respiratory maxima not linked to speech onsets ("Resp-" events) revealed significant (p(FWE) < 0.05, k > 5 cm³) bilateral responses in the postcentral sulcus, brainstem, and cerebellum. Brainstem respiratory control nuclei are known to generate automatic breathing rhythms while integrating multimodal signals for gas exchange homeostasis. The postcentral sulcus has been implicated in processing afferent feedback from respiratory effectors during intentional breathing, and the cerebellum maintains reciprocal anatomical connections with both brainstem respiratory centres and cortical respiratory control areas. Together, these structures form a network that inhibits automatic respiration for volitional respiratory control of speech onset. Our findings underscore the interaction between cortical and subcortical regions involved in respiratory control for speech production in social interactions, and offer new insights into the neural foundations of human communication and its disorders.