Human subjects were suspended in a safety harness 28 cm above the floor by a steel cable connected to a computer controlled force generator (electromagnetic brake). After the subjects were unexpectedly released, various controlled patterns of downward acceleration (less than 1 g) could be produced. During the falls, EMG activity was recorded simultaneously from the gastrocnemius, soleus, tibialis anterior, rectus femoris, and biceps femoris, along with knee and ankle joint angle in one leg. Subjects were tested eyes closed and also eyes open, both in darkness and in light using a wide field visual display. The display scene could be moved downwards at exactly the same velocity as the moving subject, left fixed with respect to the laboratory ("normal" visual field), or moved upwards at a speed equal to the subject's falling speed (upward moving visual field). Ten vestibularly normal subjects each underwent a total of 45 drops, experiencing three replications of each vision/motion combination used. Under normal visual field conditions, both short and long latency postural responses were seen, which were dependent on the magnitude of the acceleration stimulus. Several of the visual conditions significantly altered both the short and the long latency responses in most of the muscles tested. Effects were particularly prominent in the gastrocnemius and soleus, and were also more pronounced during slow (0.5 g) falls. The upward moving visual field condition increased the short latency EMG reaction in gastrocnemius and soleus for 0.5 g falls. A preliminary scheme for visual-vestibular interaction in short latency EMG responses is presented. Long latency responses are more variable and not conducive to a simple interpretation.