Hierarchical and Context-Dependent Encoding of Actions in Human Posterior Parietal and Motor Cortex

Action understanding requires internal models that link vision to motor goals. In monkeys, mirror neurons demonstrate motor resonance during observation, but single-unit evidence in humans is limited, leaving open whether such representations rely solely on motor resonance. We recorded neural activity from motor cortex (MC) and superior parietal lobule (SPL) in two tetraplegic participants implanted with Utah arrays while they intended or observed hand actions. MC strongly encoded intention but showed only weak, feature-specific overlap during observation, evident primarily at the population level. SPL, in contrast, supported shared models across intended movement and observation formats at both single-unit and population levels. In variants with incongruent instructed and observed actions, SPL encoded observed actions only when behaviorally relevant, whereas MC remained intention-dominant. Our results identify a context-dependent gating mechanism in SPL and suggest a hierarchical organization in which MC maintains intention-specific codes while SPL flexibly links observed input with internal goals to support action understanding.