Diagnostic ultrasound is used in research and clinical settings to determine the size of the lumbar spinal canal. When the technique was first introduced, measurements were obtained directly from an A-scan while concurrently viewing a B-scan. However, measurements obtained directly from a B-scan are now commonly used despite the reliability of the technique. To determine the reliability, 50 randomly selected ultrasound B-scan examinations were read on two separate occasions by three investigators. For each spinal level, the mean intraobserver error (same investigator), from the first to the second reading was determined, as was the interobserver error (between investigators) and the error due to variability in obtaining the ultrasound images. The resultant errors were approximately double those reported previously for the technique using A-scan. Interobserver variations were the major source of measurement error.