Multicategory emotions align with hierarchical event segmentation across cortical timescales

Segmenting continuous experience into discrete events supports perception, memory, and narrative understanding. While external cues have been extensively studied, the role of naturalistic emotional content is less characterized. Here, we combined high-temporal resolution annotations of 26 discrete emotion categories and a valence model with nested behavioral segmentation and fMRI during movie viewing in youth. The emotion category model, particularly sadness and disappointment, was consistently associated with neural event boundaries across the cortical hierarchy from early sensory areas to transmodal hubs in the default mode network, and aligned with behavioral segmentation at coarser temporal scales. These effects persisted after controlling for low- and high-level non-emotional features, underscoring the unique predictive value of emotion categories. Social salience cues, such as “collective presence,” also robustly predicted boundaries, suggesting complementary roles of emotional and social information. Our findings highlight the correspondence between category-specific emotional semantics and narrative structure across multiple timescales.