The concave and convex rib-vertebral angle (RVA) at levels T2-T12 was measured on AP radiographs of 19 patients with right convex idiopathic thoracic scoliosis and 10 patients with major thoracic right convex neuromuscular scoliosis. The difference between the angles on the concave and the convex sides, the RVAD, was calculated. The RVAs were also measured on radiographs from three animal groups in which spinal curves had been induced experimentally in a variety of ways. Group 1 comprised 16 rabbits that had been subjected to selective electrostimulation of the latissimus dorsi, the erector spinae and the intercostal muscles. Group 2 comprised four dead rabbits whose spines had been subjected to manual bending. Group 3 comprised eight rabbits that had undergone mechanical elongation of one rib. In both the idiopathic and the neuromuscular group, the convex RVA was smaller than the concave RVA between levels T2 and T8, with a maximal difference between T4 to T5. From T9 to T12 the concave RVA was smaller than the convex. The RVA in relation to the scoliotic segment, i.e. the apex level of the curve and the two neighbouring vertebrae above and below this level, showed similar results. With increasing Cobb angle the RVADs increased linearly with the greatest difference at the second vertebra above the apex. In the three experimental groups the pattern of the RVADs between T6 to T12 was basically similar to the findings of the clinical study. From the results of these clinical and experimental studies, it is concluded that the typical pattern of the RVAs on the concave and convex sides seems to be independent of the underlying cause of the spinal curvature. It is likely that the RVADs result from a passive mechanical adaptation of the ribs to the lateral curvature of the spine.