OBJECTIVE: To measure sagittal plane motion of lumbar vertebrae from lateral radiographic views. Previously identified factors of imprecision such as distortion in central projection, off-centre position, axial rotation, and lateral tilt of the spine were compensated. STUDY DESIGN: This study presents a new protocol to measure sagittal plane rotational and translational motion from lateral flexion-extension radiographs of the lumbar spine. BACKGROUND: Conventional methods to determine sagittal plane rotation and translation are prone to error from the distortional effects of the divergence of the radiographic beam and the measurement error inherent in constructing tangents to the contours of the vertebral body. High precision is attained by roentgen-stereophotogrammetric methods, but because of their invasive nature they can be applied only in exceptional cases. Agreement has been reached only in that measurement of sagittal plane motion from lumbar spine flexion-extension radiographs is inaccurate. Normal patterns of sagittal plane motion and the definition of what is an abnormal flexion-extension radiograph have not been settled. METHOD: Starting from an analysis of vertebral contours in the lateral view, geometric measures are identified which are virtually independent of distortion, axial rotation or lateral tilt. Applying a new protocol based on those geometric measures, the pattern of translational and rotational motion was determined from flexion-extension radiographs of 61 symptom-free, adult subjects. Measurement errors were quantified in a specimen experiment; a reproducibility study quantified inter- and intraobserver errors. RESULTS: Magnitude and sign of 'translation per degree of rotation' determined from a cohort of 61 adult subjects were very uniform for all levels of the lumbar spine. An auxiliary study evaluating a cohort of 10 healthy subjects where flexion-extension radiographs had been taken standing and side-lying showed no dependence of the rotation/translation pattern on posture. The error study demonstrated errors in angle ranging between 0.7 and 1.6 degrees and errors in displacement ranging between 1.2% and 2.4% of vertebral depth (the largest errors occurring at the L(5)/S(1) segment). Intra- and interobserver tests showed no or only negligibly small bias and an SD virtually equal to the measurement error multiplied by radical2. The relation of displacement to angle observed in the normal cohort can be used in individual cases to predict translational motion depending on the rotation actually performed. A comparison of the predicted translation (determined from normal controls) and the value actually measured allows translational hypo-, normal, or hypermobility to be quantified. Examples illustrate application of the new method in cases of normal, hypo-, and hypermobility and in the case of an instrumented spine. CONCLUSIONS: The results of this study show that precision of the measurement of rotational and translational motion can be considerably enhanced by making allowance for radiographic distortional effects and by minimizing subjective influence in the measurement procedure.