Short-timescale heart-rate fluctuation and acoustic stability show structured coupling at register transitions during singing

Heart rate (HR) is widely used to index physiological activation during singing and music performance, yet condition-level mean HR provides limited insight into how physiological regulation unfolds during vocal task execution. This exploratory study examined whether short-timescale changes in HR and acoustic stability form a structured coupling pattern at vocal register transitions, and whether that structure is shared across singers or varies across individuals and task contexts. Twenty-five trained singers completed structured vocal tasks under low-, moderate-, and high-arousal conditions. Analyses focused on Low→ Middle and Middle→ High register-transition events. Short-timescale HR change was represented by the source-to-target difference in signed maximal short-timescale HR slope (Δ s ), whereas acoustic change was represented by the source-to-target difference in the composite acoustic-stability score (Δ A ). Across 150 transition events, 72 were classified as aligned, 77 as opposed, and 1 as a tie. The aligned-minus-opposed proportion difference was −0.034, with a participant-cluster bootstrap 95% confidence interval of [−0.230, 0.160], providing no evidence for one dominant coupling direction. At the same time, the continuous association between Δ s and Δ A was weakly negative, Pearson's r = −0.180, with a participant-cluster bootstrap 95% confidence interval of [−0.317, −0.040]. Individual profiles were heterogeneous, with more than half of the singers showing mixed rather than consistently aligned or opposed coupling. These findings indicate that short-timescale HR fluctuation and acoustic stability are modestly but observably structured at register transitions, without supporting a single dominant group-level coupling direction. Physiological–acoustic coupling in singing is therefore better understood as task- and person-sensitive rather than as a uniform response or fixed physiological target. This event-based framework provides an empirical basis for person-specific regulatory profiling, multimodal performance analysis, and future human-in-the-loop computational approaches to vocal assessment and feedback.