The aim of the current study was to compare postural responses to repetitive platform-evoked perturbations in cerebellar patients with those of healthy subjects using a classical conditioning paradigm. The perturbations consisted of tilting of the platform (unconditioned stimulus: US) at random time intervals, preceded by an auditory signal that represented the conditioning stimulus (CS). Physiological reactions were recorded biomechanically by measuring the vertical ground forces, yielding the center of vertical pressure (CVP), and electrophysiologically by EMG measurements of the main muscle groups of both legs. The recording session consisted of a control section with US-alone trials, a testing section with paired stimuli and a brief final section with US-alone trials. Healthy control subjects were divided into those establishing conditioned responses (CR) in all muscles tested (strategy I) and those with CR in the gastrocnemius muscles only (strategy II), suggesting an associative motor-related process is involved. Patients with a diffuse, non-localized disease were almost unable to establish CR. This was also true for a patient with a focal surgical lesion with no CR on the affected side but who, simultaneously, showed an essentially normal CR incidence on the intact side. During US-alone trials healthy controls exhibited a remarkable decay of the UR amplitude due to a non-associative motor-related process such as habituation. The decay was most prominent in the paired trials section. In contrast, patients showed no significant differences in the UR amplitude throughout the entire recording session. Analysis of the CVP supported the electrophysiological findings, showing CR in the controls only. The differences between the responses of control subjects and those of the cerebellar patients imply strongly that the cerebellum is involved critically in controlling associative and non-associative motor-related processes.