Three-dimensional (3-D) reconstructions of the spine are being used with increasing frequency to describe scoliotic deformities, but the reproducibility of most of these techniques and the implication for the reliability of measurements made on the reconstructions has not been reported. How reliable are these reconstructions, and can a clinician interpret with confidence the results of studies based on such mathematical models? A reproducibility study of various computerised measurements obtained from 3-D reconstructions of the spine and rib cage for five subjects with adolescent idiopathic scoliosis was done to evaluate the errors associated with repeated measurements and compare them with inter- and intraobserver errors reported for similar commonly used clinical measurements. The mean variation for the Cobb angle differed according to the plane of computation from 0.6 degrees in the frontal plane to 6.7 degrees in the sagittal plane; vertebral axial rotation varied from 2.3 degrees to 5.9 degrees according to the vertebral level, and rib hump measurements displayed an average variation of 1.4 degrees. All these variations are below or within the error levels reported for equivalent 2-D measurements used by clinicians, which suggests that this 3-D model of idiopathic scoliosis may be used with confidence for clinical evaluations.