A Context-Sensitive Neural Hierarchy for Evaluating Temporal Structure in Primate Vocalizations

Understanding how the brain encodes temporal order in communication is central to explaining how complex interactions are perceived as coherent events. In humans, disrupting the sequence of words or scenes abolishes characteristic activity in higher-order networks, but whether similar mechanisms exist in nonhuman primates remains unknown. Here we used ultra-high-field fMRI (9.4 T) in awake marmosets to test how the marmoset brain evaluates temporal structure in natural conspecific vocalizations. Animals heard vocal sequences from three social contexts (angry, conversational, food-related) presented in intact, reversed, or randomized order, with call identity held constant. Disrupting sequence order altered responses across a distributed cortical-subcortical network. Contrast to reversed order, intact sequences drove stronger activation in prefrontal, cingulate, parietal, and somatosensory regions, whereas randomization produced the most widespread disruptions, additionally recruiting motor, insular, hippocampal, and thalamic territories. Uni- and multivariate analyses revealed a core network--including prefrontal area 8, cingulate areas 24/32, somatosensory cortex, and parietal Tpt--consistently sensitive to temporal coherence, with broader recruitment under severe disruption. Network-level dynamics further varied by context: conversation elicited earlier sensitivity to sequence disruptions, angry peaked later, and food built more gradually. These findings provide the first whole-brain evidence that marmosets engage hierarchically organized, context-sensitive networks to evaluate multi-agent vocal sequence structure, establishing a cross-species bridge to human narrative processing.