BACKGROUND Recent studies highlight the importance of compensatory stepping to preserve stability, and the spatial and temporal demands placed on the control of this reaction. Age-related changes in the control of stepping could greatly influence the risk of falling. The present study compares, in healthy elderly and young adults, the characteristics of compensatory stepping responses to unpredictable postural perturbations. METHODS A moving platform was used to unpredictably perturb the upright stance of 14 naive, active and mobile subjects (5 aged 22 to 28 and 9 aged 65 to 81). The first 10 randomized trials (5 forward and 5 backward) were evaluated to allow a focus on reactions to relatively novel perturbations. The behavior of the subjects was not constrained. Forceplate and kinematic measures were used to evaluate the responses evoked by the brief (600 msec) platform translation. RESULTS Subjects stepped in 98% of the trials. Although the elderly were less likely to execute a lateral anticipatory postural adjustment prior to foot-lift, the onset of swing-leg unloading tended to begin at the same time in the two age groups. There was remarkable similarity between the young and elderly in many other characteristics of the first step of the response. In spite of this similarity, the elderly subjects were twice as likely to take additional steps to regain stability (63% of trials for elderly). Moreover, in elderly subjects, the additional steps were often directed so as to preserve lateral stability, whereas the young rarely showed this tendency. CONCLUSIONS Given the functional significance of base-of-support changes as a strategy for preserving stability and the age-related differences presently revealed, assessment of the capacity to preserve stability against unpredictable perturbation, and specific measures such as the occurrence or placement of multiple steps, may prove to be a significant predictor of falling risk and an important outcome in evaluating or developing intervention strategies to prevent falls.