Differential contributions of discrete emotion categories to behavioral and neural event segmentation across...

Segmenting continuous experience into discrete events is fundamental for perception, memory yet the contribution of fine-grained emotion categories to this process remains unclear. Using frame-level annotations of 26 discrete emotion categories, valence, and transient emotional fluctuations, we examined how emotion categories predict behavioral and neural event segmentation during movie viewing. Across hierarchical (event-nesting) behavioral tasks, category-based models, particularly sadness and disappointment, more robustly predicted event boundaries than valence-and-intensity-based models, indicating a unique contribution of emotion semantics to narrative structure. In fMRI, these same emotion categories reliably predicted neural boundaries derived from Hidden Markov Model (HMM) across the cortical hierarchy, from early sensory regions to default-mode hubs. Transient, cosine-based fluctuations of overall emotion semantics additionally predicted coarse-scale boundaries in nucleus accumbens (NAcc). Social salience cues, especially collective presence provided complementary predictive power. Together, these findings indicate that discrete emotion semantics and their dynamics provide cross-timescale signals that support both behavioral and neural event segmentation, highlighting the contribution of emotion to the organization of continuous experience.