Exposure to laterally displacing prisms is characterized by systematic misreaching in the opposite direction after prisms are removed. Other learning tasks involving altered visuomotor mappings can often be mastered by the subject with minimal resulting aftereffects. One variable that may account for this difference is the nature of the feedback provided to the subject: during studies of prism exposure, subjects usually view the hand itself, whereas in many studies of visuomotor learning, subjects view a computer-generated representation of the hand position or movement. We compared the use of actual feedback of the hand with computer-generated representational feedback of its position during exposure to laterally displacing prisms. In the actual feedback condition (ACT), a light on the fingertip was illuminated immediately at the end of each reach. In the representational feedback condition (REP), a computer-generated spot of light was displayed to indicate the exact position of the fingertip at the end of each reach. Whereas the rate and magnitude of error correction were the same in both conditions, only the ACT condition produced the large adaptive aftereffect typically observed after prism exposure. These results suggest that the perception of a physical coincidence between the feedback source and the hand may be a key factor in determining whether adaptation is accomplished through perceptual recalibration or visuomotor skill acquisition.