Although vestibular input codes head acceleration, it is not clear whether or not this signal is critical for triggering the initial postural response to a perturbation of stance, and for determining the appropriate direction of response. These experiments were designed to examine the contribution of vestibular inputs to the control of balance in the freely standing cat. Four cats were trained to stand quietly on a moveable force platform. The animal's stance was unexpected perturbed by applying a linear ramp-and-hold translation to the support surface in each of eight different directions in the horizontal plane. The characteristics of quiet stance and the response to the perturbations were quantified in terms of the 3-D ground reaction forces under each paw and the EMG activity in selected muscles. The animals were bilaterally labyrinthectomized, and their responses compared before and after lesion. The cats were able to stand stably on the platform within 2-3 days of the lesion. During quiet stance, there was no change in the distribution of vertical forces under the paws and no increase in sway area. Horizontal plane forces, which were normally outwardly directed on the diagonals, became more laterally directed and transiently larger in amplitude. The level of tonic EMG activity increased in some extensors and flexors, and decreased in others, compared to control. The responses to platform translation were characterized by normal spatial and temporal patterns and latencies of EMG activity. Furthermore, all cats continued to use the force constraint strategy that is characteristic of the intact animal (Macpherson, 1988a). The only clear deficit in performance was a transient hypermetria, characterized by an over-response to the translation. Although the cats over-responded, they were still able to maintain their balance successfully. The moderate changes in quiet stance and in response to perturbation gradually returned to control values over 8-10 days following the lesion. These results suggest that vestibular information is not necessary for triggering appropriate postural responses evoked by support surface translations, nor for selecting the direction of response.