Learning Environmental Dynamics: Building Internal Models of a Simulated External Force

Motor adaptation is typically studied using limb-based perturbations in highly constrained tasks. However, everyday actions require compensating for uncertain environmental dynamics impacting manipulated objects within redundant execution spaces, where multiple combinations of variables can achieve success. To investigate this, we designed a novel virtual reality task where participants launched a ball across a lateral water current to spatially distinct targets. Two experiments (2.0 m/s and 3.0 m/s) tested if classic adaptation signatures remain consistent across varying perturbation sizes. Results revealed three key signatures. First, participants exhibited robust error reduction to all targets during training. Second, between-subjects analyses revealed that this learning produced substantial, immediate transfer to untrained target locations. Finally, persistent motor aftereffects appeared upon perturbation removal, despite explicit cues indicating absence of the perturbation. Together, these results demonstrate that humans form internal models under altered projectile-environment dynamics, and that prior experience facilitates broad generalization across the workspace.