Abstract
We developed an embodied Virtual Reality (eVR) environment and paradigm for studying sensorimotor learning mechanisms in real-world tasks. By using a real-world setting, i.e. playing pool billiard, we showed previously that subjects could learn by error-based and reward-based learning mechanisms, yet actually split into two distinct groups using one or the other mechanism primarily. Now, we isolated Error-based and Reward-based reinforcement learning by providing different partial visual feedback to two groups of subjects (that clamped either error or reward feedback), to understand how and whether the learning processes were different between groups. Visual feedback was manipulated in form of a rotation of the target ball’s trajectory, that requires from learners a cognitive abstraction that acts only on an object, in contrast to traditional visuomotor learning experiments which rotate the whole visual field. We found that subjects’ compensation for the perturbation and the learning curve differed between groups, with the error-based subjects showing exponential learning curve and achieving a higher improvement than reward subjects who presented a linear learning curve. Yet, no significant difference was found in inter-trial variability and success rate.