A large body of evidence has been collected which describes the response parameters associated with automatic balance corrections in man to perturbations in the pitch plane. However, perturbations to human stance can be expected from multiple directions. The purpose of the present study was to describe the directional sensitivities of muscle responses re-establishing disturbed stance equilibrium in normal subjects. The contributions of stretch reflex and automatic balance-correcting responses to balance control, and concomitant biomechanical reactions, were examined for combinations of pitch and roll perturbations of the support surface. More specifically, muscle responses, initial head accelerations and trunk velocities were analyzed with the intention of identifying possible origins of directionally specific triggering signals and to examine how sensory information is used to modulate triggered balance corrections with respect to direction. Fourteen healthy adults were required to stand on a dual-axis rotating platform capable of delivering rotational perturbations with constant amplitude (7.5 degrees ) and velocity (50 degrees /s) through multiple directions in the pitch and roll planes. Each subject was randomly presented with 44 support surface rotations through 16 different directions separated by 22.5 degrees first under eyes-open, and then, for a second identical set of rotations, under eyes-closed conditions. Bilateral muscle activities from tibialis anterior, soleus, lateral quadriceps and paraspinals were recorded, averaged across direction, and areas calculated over intervals with significant bursts of activity. Trunk angular velocity and ankle torque data were averaged over intervals corresponding to significant biomechanical events. Stretch reflex (intervals of 40-100, 80-120 ms) and automatic balance-correcting responses (120-220, 240-340 ms) in the same muscle were sensitive to distinctly different directions. The directions of the maximum amplitude of balance-correcting activity in leg muscles were oriented along the pitch plane, approximately 180 degrees from the maximum amplitude of stretch responses. Ankle torques for almost all perturbation directions were also aligned along the pitch plane. Stretch reflexes in paraspinal muscles were tuned along the 45 degrees plane but at 90 degrees to automatic balance corrections and 180 degrees to unloading responses in the same muscle. Stretch reflex onsets in paraspinal muscles were observed at 60 ms, as early as those of soleus muscles. In contrast, unloading reflexes in released paraspinal muscles were observed at 40 ms for perturbations which caused roll of the trunk towards the recorded muscle. Onsets of trunk roll velocities were earlier and more rapid than those observed for pitch velocities. Trunk pitch occurred for pure roll directions but not vice versa. When considered together, early stretch and unloading of paraspinals, and concomitant roll and pitch velocities of the trunk requiring a roll-and-pitch-based hip torque strategy, bring into question previous hypotheses of an ankle-based trigger signal or ankle-based movement strategies for postural balance reactions. These findings are compatible with the hypothesis that stretch-, force- and joint-related proprioceptive receptors at the level of the trunk provide a directionally sensitive triggering mechanism underlying a, minimally, two-stage (pitch-based leg and pitch-and-roll-based trunk) balance-correcting strategy. Accelerometer recordings from the head identified large vertical linear accelerations only for pitch movements and angular roll accelerations only during roll perturbations with latencies as early as 15 ms. Thus, it appears that balance corrections in leg and trunk muscles may receive strong, receptor-dependent (otolith or vertical canal) and directionally sensitive amplitude-modulating input from vestibulospinal signals.