Alignment in Action Anticipation is Key to Successful Coordination in Dynamic Interactive Tasks

The success and failure of human coordination critically depend on the ability to anticipate and adapt to both one’s own and others’ actions. While theoretical models emphasize the central role of predictive mechanisms, empirical studies often rely on simplified tasks that do not capture the sensory, motor, and social complexity of real-world joint action. Consequently, it remains unclear which among the many behavioral, cognitive, physiological, demographic, and relational factors truly contribute to coordination success.Here, we introduce a quantitative framework to address this gap by identifying key predictors of performance in a fast-paced ball-hitting task, performed both individually (individual-action) and in a turn-taking format with a partner (joint-action). Using high-resolution multimodal tracking of eye and body movements and heart rate, combined with machine learning and Bayesian modeling, we show that coordination success is primarily associated with the similarity in predictive styles between partners, rather than with general anticipatory ability alone. Well-coordinated pairs—defined by sustained, accurate ball exchanges—exhibited similar timing and patterns of anticipatory saccades in the individual-action task and were more synchronized in their predictions of both their own and their partner’s actions during joint performance. In contrast, demographic and relational variables (e.g., gender, height, familiarity), as well as global physiological and motor activity (e.g., overall movement, heart rate), showed minimal association with coordination outcomes.By reverse-engineering the behavioral signatures of successful coordination, our study highlights alignment in action anticipation as a key mechanism and offers a data-driven foundation for improving joint performance in both human–human and human–robot interaction.