Human motor learning affects not only motion pattern but also perception. On the basis of this idea, we investigated some human-balancing tasks to observe changes in the balance perception. We simulated one of the results to replay the human behavior. In this study, we aim to demonstrate the human adaptive behavior in a motor and sensory system using a robot in a real-world scenario. The subjective upright, i.e., the inclination angle was considered as upright, was evaluated as perception in a balance position, and its lateral shift was determined after a motor-learning task under lateral disturbances. The adaptation dynamics was defined based on our hypothesis that stated that the subjective upright tends to vary toward the posture in which balance can be best maintained. Consequently, the change in the subjective upright similar to the human result was reproduced using a two-link robot that was subjected to the same environmental condition as that in the human experiment.