The Computational Grammar of Human Behaviour in a Biological Environment

DynaClade segments human full-body motion into discrete actions and reveals a previously undescribed pattern—greater body sway without locomotion—in VR risky foraging.

• Why it matters: Action organization between abstract goals and movement execution is poorly characterized, limiting how we understand how environmental affordances shape spontaneous behavior.
• What they did: The unsupervised temporal segmentation algorithm DynaClade was applied to freely moving participants’ full-body motion capture in ecologically realistic risky foraging virtual reality.
• The result: Action sequences were highly non-random with history dependence beyond first-order Markov structure, largely explained by second-order prediction and captured by a hidden Markov model with context-specific latent states.

AI summary of the abstract · may contain errors · not medical advice · read the paper.