<p>Simulated environments, e.g., virtual or augmented reality environments, are becoming increasingly popular for the investigation and training of motor actions. Yet, so far it remains unclear if results of research and training in those environments transfer in the expected way to natural environments. Here, we investigated the types of visual cues that are required to ensure naturalistic hand movements in simulated environments. We compared obstacle avoidance of physical objects with obstacle avoidance of closely matched 2D and 3D images of the physical objects. Participants were asked to reach towards a target position without colliding with obstacles of varying height that were placed in the movement path. Using a pre-test post-test design, we tested obstacle avoidance for 2D and 3D images of obstacles both before and after exposure to the physical obstacles. Consistent with previous findings, we found that participants initially underestimated the magnitude differences between the obstacles, but after exposure to the physical obstacles avoidance performance for the 3D images became similar to performance for the physical obstacles. No such change was found for 2D images. Our findings highlight the importance of disparity cues for naturalistic motor actions in personal space. Furthermore, they suggest that the observed change in obstacle avoidance for 3D images resulted from a calibration of the disparity cues in the 3D images using an accurate estimate of the egocentric distance to the obstacles gained from the interaction with the physical obstacles.</p>