Heart rate variability dynamics reveal mechanisms for music-based autonomic modulation

Background Music is increasingly recognised as a non-pharmacological modulator of cardiovascular autonomic function [1]. However, it remains unclear which musical features drive these effects, whether music alters mean autonomic tone or variability, and whether changes persist beyond listening. Purpose To quantify feature-specific autonomic responses to music and characterise acute and post-listening modulation of heart rate variability (HRV), a clinically relevant marker of vagal function and cardiovascular risk. Methods RR data were analysed from the HeartFM study [2], in which 126 participants (74 women; mean 43.3 years [95%CI 40.5–46.0]) listened to 9 piano tracks selected from a pool of 30 (Fig.2A). Of these, 34 had complete baseline and post-listening (endline) recordings. Time-domain (RMSSD, SDNN) and frequency-domain HRV metrics (LF and HF power, two spectral methods) were computed in 60s windows. Track-level analyses examined within-track HRV mean (HRV_mean) and standard deviation (HRV_std). Musical predictors included tempo and loudness, within-track acoustic spread, and phrase-arc structure (mean length and variability). Significance was assessed using ANOVA/Kruskal–Wallis, correlation analyses with FDR correction, and effect size with Cohen’s d. Results No musical feature influenced HRV_mean at the track level. HRV_std, however, was strongly modulated by music. Tempo showed a stepwise suppression of HRV_std across all indices (η²=0.052–0.069; q<10⁻¹⁹), with faster tracks producing progressively lower within-track variability (RMSSD_std track-level r=−0.81; 111/126 participants negative, p<10⁻¹⁹). Within-track tempo std predicted lower HRV_std (ρ=−0.24; 109/126 negative), suggesting suppressive mechanism of wider tempo deviations. Loudness showed a weaker effect driven exclusively by the loudest tracks (η²≈0.01–0.02), with medium and soft levels statistically indistinguishable. Greater loudness phrase arc length and variability suppressed HRV_std. See Fig.1. At session level, HRV_mean did not shift during music, but autonomic fluctuation increased (SDNN_std d=0.93; LF_std d=0.56; RMSSD_std d=0.46), reflecting dynamic engagement rather than a steady-state shift. Endline HRV_mean exceeded baseline (RMSSD +22%, SDNN +20%, both q<0.05) with no regression-to-mean effect, consistent with parasympathetic upregulation. Spectral HRV_mean power changes were not significant after FDR correction. HRV_std elevation persists at endline (SDNN_std p<0.021) while spectral variability normalised. See Fig.2. Conclusions Music modulates autonomic dynamics by altering variability during listening and enhancing vagal tone after exposure. These findings suggest music may "exercise" autonomic flexibility, potentially through entrainment [3], rather than simply induce relaxation. This has implications for the implementation of music-based interventions targeting autonomic flexibility in cardiovascular prevention and recovery.