1. Electromyographic (EMG) responses were recorded in both legs, along with corresponding joint movements, after uni- and bilateral perturbations during stance on a treadmill with split belts. Displacements were directed forward, backward, or in opposing directions. They were induced by randomly timed ramp impulses at one of four different rates of treadmill acceleration. 2. Unilateral perturbations directed backward were followed by a bilateral gastrocnemius-EMG response, forward-directed perturbations by a bilateral tibialis anterior-EMG response. The amplitude of these responses was dependent on the rate of treadmill acceleration. Relative to the response of the displaced leg, the amplitude of the EMG response on the nondisplaced side was smaller when a gastrocnemius EMG response was induced, and about equal when the tibialis anterior muscle was activated. The onset latencies were shorter on the displaced side (displaced leg 75-96 ms, non-displaced leg 93-112 ms). 3. Bilateral perturbations in one direction were followed by larger EMG responses in both legs (in the gastrocnemius for backward-directed impulses, in the tibialis anterior for forward-directed impulses). For a given acceleration rate, their amplitude was about equal to the sum of the EMG amplitude of the displaced leg and that of the nondisplaced leg obtained during unilateral displacement. The inverse result was obtained when the legs were simultaneously displaced in opposite directions: EMG responses in both legs were significantly smaller than those obtained after unilateral displacement. 4. It is concluded that a unilateral displacement evokes reflex EMG responses in the synergistic muscles of both legs, which are graded according to the size of the proprioceptive input from the primarily displaced joint. During bilateral displacements, the activity induced by the respective contralateral leg is linearly summed or subtracted, depending on whether the legs are displaced in the same or in opposite directions. In view of the short latencies of these bilateral responses, it would seem that they are mediated by a spinal mechanism. 5. Distinct differences in the behavior of the antagonistic leg muscles were observed: 1) the coactivation of the contralateral leg muscle was significantly smaller when the gastrocnemius was stretched unilaterally, whereas it was about equal for the tibialis anterior; and 2) the gastrocnemius EMG responses were closely correlated with the displacement velocity, whereas the tibialis anterior response was more closely correlated with acceleration, i.e., the tibialis anterior response was more dynamic in nature.(ABSTRACT TRUNCATED AT 400 WORDS)