Objectives. To compare the Cobb technique for measuring kyphosis with an alternative Cobb method and a computer-assisted curve assessment technique, and to examine the influence of vertebral body and disc shape on kyphosis.Methods. Kyphosis measurements were derived from 93 lateral spinal radiographs or sagittal computed tomography images of cadaveric spines, using: (i) a computer-assisted method for estimating radius of curvature; (ii) the traditional Cobb method and (iii) an alternative Cobb method. Regression models were applied for agreement analyses, and to examine the relative contribution of vertebral body and disc shape on the extent of curvature.Results and conclusions. Strong associations existed between curvature and angle data derived from the three methods. confirming the clinical utility of these techniques for the quantification of thoracic kyphosis. However, the traditional Cobb method tended to overestimate kyphosis in the presence of vertebral body end-plate deformation. The degree of kyphosis was strongly reflective of the extent of deformity of the vertebral bodies, and to a lesser extent the shape of the thoracic discs.