We have developed a novel VR task: the Dynamic Reaching Test, that measures human forearm movement in 3D space. In this task, three different stereoscopic displays: autostereoscopic (AS), shutter glasses (SG) and head mounted display (HMD), are used in tests in which subjects must catch a virtual ball thrown at them. Parameters such as percentage of successful catches, movement efficiency (subject path length compared to minimal path length), and reaction time are measured to evaluate differences in 3D perception among the three stereoscopic displays. The SG produces the highest percentage of successful catches, though the difference between the three displays is small, implying that users can perform the VR task with any of the displays. The SG and HMD produced the best movement efficiency, while the AS was slightly less efficient. Finally, the AS and HMD produced similar reaction times that were slightly higher (by 0.1 s) than the SG. We conclude that SG and HMD displays were the most effective, but only slightly better than the AS display.