Human escape in freely moving virtual reality follows a structured movement pattern shaped by threat and context

Evading danger is critical to survival. In non-human animals, escape strategies are shaped by neural, biomechanical, and ecological constraints, resulting in species-specific patterns. In humans, ethical and practical constraints have until recently hindered investigation of escape movements, such that its organising principles are commonly extrapolated from other, mostly quadruped, species. Here, we use freely moving virtual reality (fm-VR) in a large physical space to present biologically relevant threats. We discover that human escape behaviour is organized within a constrained action space shaped by threat and context, revealing a small set of stereotyped kinematic patterns that are not predicted from those reported in other mammals. Pattern selection is shaped by environment and individual preference, and is not predicted by behaviour in safe conditions. The dominant kinematic pattern includes head orientation toward the threat, body rotation beyond this angle until facing away, and escape initiation with ipsilateral foot. Alternative variants include turning away from the threat, backward movement, and misdirected flight. Certain alternative escape patterns and kinematic features reduce success, whereas specific preparatory adjustments enhance it. Our results provide a foundation for probing the neural mechanisms of human escape, and enable investigation of potential disruptions in clinical conditions.