Individuals’ Acoustic Features and Heart Rate Patterns Reveal Team Differences in High-Stakes Collaborative Learning

Abstract

Effective collaboration in high-stakes learning environments, such as nursing simulations, relies not only on verbal communication but also on internal states (e.g., stress and engagement), often reflected in both speech and physiological data. However, the relationship between acoustic speech features and physiological arousal remains underexplored in authentic, team-based scenarios. This study investigates how speech acoustics correlate with heart rate relative to baseline (HRrelative) during simulation-based learning in healthcare education. We analysed speech and physiological data from 50 team sessions (173 students), extracting GeMAPS features and aligning them with utterance-level HR data. Correlation and predictive models were applied across simulation phases and team performance levels. Results reveal that high-performing teams modulated speech features—such as pitch, articulation, and loudness—more consistently across phases, suggesting adaptive regulation under pressure. On the contrary, low-performing teams showed similar shifts, but with less structure and at later phases. These findings demonstrate the potential of multimodal data to reveal hidden patterns in teamwork. Speech–physiology dynamics could inform targeted feedback during debriefing, supporting communication and leadership training in healthcare education.