A balance testing methodology has been developed based on a posture control model which defines relative stability by the degree to which a transient postural perturbation causes the center-of-pressure on the feet to approach the limits of the base-of-support. To minimize anticipatory adaptations and to ensure subject safety, the balance test uses a small-amplitude continuous random or pseudo-random perturbation. The data are used to identify an input-output model, which is then used to predict large-amplitude transient response. Sixtyfour normal subjects were tested, as well as five patients with vestibular lesions and five elderly subjects with a history of falling. Tests were performed with eyes open (EO) and blindfolded (BF), and the EO/BF ratio was calculated. The results showed highly significant differences between young (20-40) and elderly (61-79) normals, but no differences between males and females. In using the balance test to identify the balance-impaired individuals, the BF and EO/BF data provided higher success rates than the EO data. Depending on the modeling method used, the balance test was able to identify up to three of five vestibular patients and five of five elderly fallers, at a false positive rate of 25 percent in the normal subjects.