1. To identify the vestibular contribution to human standing, responses in leg muscles evoked by galvanic vestibular stimulation were studied. Step impulses of current were applied between the mastoid processes of normal subjects and the effects on the soleus and tibialis anterior electromyograms (EMGs), ankle torque, and body sway were identified by post-stimulus averaging. The responses were measured when subjects stood on a stable platform or on an unstable platform and the effects of eye closure were also assessed. Responses were also recorded during voluntary contraction of the leg muscles and when subjects balanced a load equivalent to their own body in a situation where vestibular postural reflexes would not be useful. 2. At a mean post-stimulus latency of 56 ms, there were reciprocal changes in soleus and tibialis anterior muscle activity followed, at 105 ms, by larger responses of opposite sign. These were termed the short- and middle-latency responses, respectively. Both responses increased with stimulus intensity, but the short-latency response had a higher threshold. The early response had a similar latency to EMG responses evoked by rapid postural perturbations. Both responses were larger when the eyes were closed, but eye closure was associated with increased sway and EMG activity, and the responses were of similar magnitude when scaled to background EMG level. 3. Both short- and middle-latency EMG responses in soleus and tibialis anterior muscles produced small transient postural sways. The transient changes in EMG activity were followed by a larger prolonged sway which was not attributable to the activity in these muscles but rather to reflex or volitional adjustments to movements at other body segments. When subjects were prevented from swaying, the galvanic stimulus produced illusory movements in the opposite direction to the sway evoked when standing, and it is possible that the prolonged sway is a reaction to the illusion of sway. 4. The short- and middle-latency responses were modified during different postural tasks according to the dependence on vestibular reflexes. When the support platform was unstable, the EMG responses to galvanic stimulation were larger. There were no vestibular-evoked responses when seated subjects made voluntary contractions of the leg muscles or when they stood upright with the trunk supported, using the ankles to balance a body-like load.(ABSTRACT TRUNCATED AT 400 WORDS)