Axial vertebral rotation (AVR) of normal and scoliotic vertebrae was measured in computed tomography (CT) images by three observers using different manual and a computerized method. To analyze 4 manual and a computerized method for measuring AVR in CT images. Manual measurement of AVR is difficult and error-prone when compared with computerized methods. To the best of our knowledge, a systematic comparison of the established manual with a computerized method has not been performed yet. AVRs of 14 normal and 14 scoliotic vertebrae from CT images were measured (a) manually using the methods of Aaro and Dahlborn, Spine 1981;6:460-7, Ho et al, Spine 1993;18:1173-7, Krismer et al, J Spinal Disord 1999;12:126-30, and Göçen et al, J Spinal Disord 1998;11:210-4., and (b) automatically using a computerized method, which is based on the evaluation of vertebral symmetry in 2 dimensions (2D) and in 3 dimensions (3D). RESULTS.: The computerized method was most consistent with the method of Aaro and Dahlborn, which also proved to be the most reproducible and reliable manual method. The low overall intraobserver variability (1.1 degrees SD) and interobserver variability (1.8 degrees SD) of the computerized method indicate that the symmetry-based determination of AVR is reproducible and reliable, as the localization of vertebral centroids is the major source of its variability. The computerized method yielded higher reproducibility and reliability of AVR measurements, indicating that it may represent a feasible alternative to manual methods, moreover because it is also faster and more operator-friendly.