The team analyzed both dimensional emotions, captured as valence and arousal, and seven discrete emotions: anger, gentleness, happiness, peacefulness, sadness, solemnness, and transcendence. The linear models showed that BCSC features explained 35.1 percent of the variance in perceived valence and 37.1 percent in arousal. These are substantial proportions for purely structural acoustic and notational features, indicating that listeners appear to read emotion from the architecture of a piece rather than from cultural labels alone.

Specific predictors stood out. For valence, Total Asymmetry, Average Local First-Order Entropy, and Third-Order Duration Entropy were among the strongest positive predictors, while Event Density showed a strong negative effect. In other words, pieces perceived as more pleasant tended to be more asymmetric and locally unpredictable in their pitch and duration patterns, but less densely packed with events. For arousal, Duration Abruptness, Total Asymmetry, Average Local First-Order Entropy, and Third-Order Duration Entropy were significant positive predictors, whereas Rhythmic Abruptness, Event Density, and First-Order Duration Entropy had significant negative effects.

The discrete-emotion models revealed even finer structure. Predictive power ranged from an adjusted R-squared of 0.157 for peacefulness to 0.386 for happiness. The same feature could push two emotions in opposite directions: First-Order Duration Entropy was negatively associated with perceived anger but positively associated with peacefulness. Bi-Unbalance, a measure of structural imbalance, predicted anger, suggesting that listeners associate lopsided musical organization with tension and aggression.

The nonlinear random forest analyses confirmed the linear findings, highlighting temporal irregularity, melodic interval size, and structural imbalance as the most informative cues. The convergence of two very different modeling approaches strengthens the conclusion that these relationships are not statistical artifacts but reflect genuine regularities in how musical structure signals emotion.

The authors frame these results as evidence that mid-level structural features provide systematic emotional cues in Chinese traditional music, a repertoire whose structural-emotional mapping has been underexplored compared with Western classical and pop music. Because the features are computable from scores or audio, the framework could inform music information retrieval systems, emotion-aware playlist generation, and computational analyses of traditional repertoires.

The study also carries limitations worth noting. The reported relationships are predictive associations between structural features and perceived emotions, not demonstrations of causal mechanisms, and the findings concern perception rather than the felt emotions of listeners or performers. The variance explained, while meaningful, leaves most of the emotional signal to other factors such as timbre, lyrics, performance style, and listener background that were not the focus of the BCSC framework.

Nevertheless, the work offers a concrete vocabulary for discussing emotion in Chinese traditional music. Terms like entropy, asymmetry, abruptness, and event density translate subjective impressions—flowing, agitated, solemn—into quantifiable properties, opening a path toward cross-cultural comparisons of how musical structure encodes feeling.

Subject of Research: Structural predictors of perceived emotion in Chinese traditional music

Article Title: How balance, contour, symmetry, and complexity predict the perceived emotions in Chinese traditional music

Article References: Shen, H., Liu, Y., Li, J., Wu, Y., Chao, C., & Xu, L. (2026). How balance, contour, symmetry, and complexity predict the perceived emotions in Chinese traditional music. Humanities and Social Sciences Communications. https://doi.org/10.1057/s41599-026-09611-9

Image Credits: AI Generated

DOI: 10.1057/s41599-026-09611-9

Keywords: music psychology, Chinese traditional music, emotion perception, BCSC framework, valence, arousal, regression analysis, random forest, musical symmetry, entropy, music information retrieval, balance