The use of flexicurves to measure lumbar sagittal mobility was subjected to a series of reliability and validation experiments. Appropriate statistical methods were described and used to quantify intraobserver and intrasubject variability and to determine limits of agreement with measurements from radiographs. It was shown that the traditional use of correlation coefficients can produce misleading or inadequate information. The flexicurve technique had an intraobserver variability of 3-4 degrees of movement, was not significantly influenced by intrasubject variability, and provided measurements typically within 6 degrees of radiographic measurements. The data suggest that the flexicurve technique is less biased than the inclinometric method. These results demonstrate the use of suitable statistical methods to assess the clinical usefulness, or level of interchangeability, of spinal measurement instruments.