The representation of speech conversations in the human auditory cortex: role of social and semantic factors

Human social nature has shaped auditory perception, as hearing is essential for navigating social interactions, especially when listening to others' conversations. While much research has examined how the brain processes isolated words or sentences, far less is known about how broader social and semantic contexts influence speech comprehension. We used 7 T fMRI to examine neural responses while participants listened to two-speaker dialogues versus single-speaker monologues, presented either in intact or sentence-scrambled order. Twenty-four healthy young adults listened to AI-generated five-sentence conversations designed to independently manipulate social (dialogue vs. monologue) and semantic (intact vs. sentence-scrambled) contexts. Whole-brain univariate analyses revealed increased activity for scrambled compared to intact conversations in the left superior temporal sulcus (STS), consistent with predictive-coding models. Although social context alone showed no main effect, an interaction emerged: semantic disruption elicited stronger responses in dialogues than monologues within the STS. Multivariate pattern analyses further revealed higher classification accuracy of individual sentences within dialogues vs. monologues, particularly in the left anterior STS and inferior frontal gyrus, suggesting that social context enhances linguistic encoding. Together, these findings indicate that the left STS integrates both semantic and social information, supporting predictive and context-sensitive mechanisms crucial for real-world verbal communication.