1. The effect of central set on automatic postural responses was studied in humans exposed to horizontal support-surface perturbations causing forward sway. Central set was varied by providing subjects with prior experience of postural stimulus velocities or amplitudes under 1) serial and random conditions, 2) expected and unexpected conditions, and 3) practiced and unpracticed conditions. In particular, the influence of central-set conditions was examined on the pattern and magnitude of six leg and trunk electromyograph (EMG) activations and associated ankle torque responses to postural perturbations with identical stimulus parameters. 2. The scaling of initial agonist integrated EMG (IEMG) and torque responses to postural perturbation amplitude disappeared when perturbation amplitudes were randomized. This finding suggests that the initial magnitude of postural responses were centrally set to anticipated postural perturbation amplitudes based on sequential experience with the stimulus. 3. Expectation of postural stimulus amplitude had a significant effect on initial torque responses; subjects overresponded when a larger perturbation was expected and underresponded when a smaller perturbation was expected. Expectation of postural stimulus velocity had a smaller effect on initial torque responses, and subjects consistently overresponded when the velocity of the perturbation was unexpected. This difference in amplitude and velocity expectation may be because of the capacity to encode stimulus velocity, but not amplitude information, into the earliest postural responses of the current trial. The relative strength of amplitude and velocity central-set effects varied widely with individual subjects. 4. Central-set conditions did not affect initial EMG response latencies (100 +/- 20 ms, mean +/- SD) or the relative onset of proximal and distal agonists and antagonists. Unexpected or unpracticed stimulus amplitudes, however, were associated with significant late activation of ankle antagonist, tibialis. Thus errors in initial response magnitude because of central-set effects appear to be partially corrected by reciprocal antagonist activity. Agonist IEMG, however, did not always reflect significant changes in torque responses with central-set conditions. 5. Expectation of postural stimulus amplitude and velocity had two different types of effects on the magnitudes of postural responses: 1) a directionally specific, central-set effect consisting of either increased or decreased responses, depending on expectation of stimulus amplitude; and 2) a nonspecific enhanced response to novel stimulus velocities with a gradual reduction when a velocity was presented repeatedly. Two different neural mechanisms are proposed for these two adaptive effects. 6. Reduction of postural response magnitude and antagonist activity during practice may be partially explained by adaptive mechanisms based on expectation because of prior experience with stimulus velocity and amp