Lumbar spinal flexion and extension motion was measured radiographically and from the back surface curvature in 42 patients with low-back pain to determine whether the surface measurements provided accurate measurement of the total lumbar spinal motion and its distribution by vertebral level. Also, a biomechanical model of motion of the lumbar spine was used to predict the changes in the back surface as a function of varying amount of flexion and extension at the intervertebral articulations. Surface measurements provided reasonably accurate (+/- 11.2 degrees or +/- 25.5%) measurement of the total lumbar motion with a correlation coefficient of 0.58 between surface and radiographic measures. However, there was a poor correlation (varying between -.08 and .54 at different levels) between segmental motion measured radiographically and at the surface. Detection of the most mobile and least mobile intervertebral levels was done successfully by the surface curvature method in no more cases than would be expected by chance. Measurements of intersegmental motion are inherently error prone since they involve calculating differences between small angles which in turn are difficult to measure accurately. Surface measurements based on changes in back curvature are further complicated since the back surface has a variable relationship with spine shape and accurate measurement of curvature is very difficult.