This study was designed to provide evidence for the hypothesis that human balance corrections in response to pitch perturbations are controlled by muscle action mainly about the ankle and knee joints, while balance corrections for roll perturbations are controlled predominantly by motion about the hip and lumbro-sacral joints, with limited interaction between muscle action at the ankle and trunk even when the perturbation is multi-directional. A dual-axis rotating support surface delivered unexpected random perturbations to the stance of 19 healthy young adults through 8 different directions in the pitch and the roll planes, and 3 delays between pitch and roll directions. Each rotational perturbation had a constant total amplitude of 7.5 deg and an angular velocity of 60 deg/sec. Roll-delays with respect to pitch were: no delay (roll and pitch movements of the support-surface occurred simultaneously), 'short delay' (a 50 ms delay of roll with respect to pitch movements, chosen to correspond to the onset-time of leg-muscle stretch-reflexes), and 'long delay' (a 150 ms delay between roll and pitch movements chosen to shift the time when trunk roll velocity peaks to the time when trunk peak pitch velocity normally occurs). Biomechanical measures included pitch and roll measurements at different body segments and ground-reaction forces. EMG activity was recorded from several ankle and trunk muscles. Delays of stimulus roll with respect to pitch resulted in significantly delayed biomechanical roll responses of the legs, trunk, arms and head without any significant changes in roll velocity amplitude with respect to the no delay condition. Peak roll velocities of these segments occurred at times consistent with the stimulus delay. Delayed roll perturbations induced only small changes in the pitch motion of the legs and trunk as well as ankle torques, however major changes in the time when roll motion of the trunk was arrested. Amplitudes and directional sensitivity of short-latency (SL) stretch reflexes in ankle muscles were not altered with increasing roll delay. Small changes to balance correcting responses in ankle muscles were observed. SL stretch reflexes in hip and trunk muscles were delayed and balance- correcting responses in trunk muscles became split into two distinct responses with delayed roll. The first of these responses was small and had a directional responsiveness aligned more along the pitch plane. The main, larger, response occurred with an onset and time-to-peak consistent with the delay in trunk roll displacement and its directional responsiveness was roll oriented. The sum of the amplitudes of these two types of balance-correcting responses remained constant with roll delay. These results support the hypothesis that corrections of the body's pitch and roll motion are programmed separately by the neural command signals, and provide insights into possible triggering mechanisms. The fact that SL stretch reflexes were significantly delayed with delayed roll in hip and trunk muscles but lower-leg balance-correcting responses were altered minimally suggests that balance-corrections in the pitch plane are not dependent on a triggering action from upper body, proprioceptive input. The evidence that lower-leg muscle balance-correcting activity is hardly changed by delayed trunk roll also indicates that lower-leg muscle activity is not predominant in correcting roll motion of the body. On the other hand, the delayed balance-correcting activity in trunk muscles, despite early onset of SL stretch reflexes in ankle muscles, suggests that roll-directed balance-correcting activity in trunk muscles is triggered by afferent inputs also used to generate SL stretch reflex activity in trunk muscles and not by lower-leg afferent information. Lower-leg and trunk muscle activity appears to have a dual action in balance corrections. In trunk muscles the main action is to correct for roll perturbations and the lesser action may be an anticipatory stabilising reaction for pitch perturbations. Likewise the small changes in lower-leg muscle activity may result from a generalised stabilising reaction to roll perturbations, but the main action is to correct for pitch perturbations.