OBJECTIVE Upright flexion-extension myelography has shown marked variation in the severity of lumbar dural sac stenosis during flexion and extension in some patients ("positionally dependent" stenosis). Axial computed tomography (CT) of the lumbar spine while the patient is supine may not demonstrate the maximal spinal stenosis shown by flexion-extension myelography. This study compares the severity of central lumbar spinal stenosis on CT myelograms and on upright flexion-extension myelograms in patients with positionally dependent spinal stenosis. METHODS From January 1995 to December 1996, 210 patients underwent upright flexion-extension lumbar myelography for the assessment of back pain, radiculopathy or neurogenic claudication. In 33 of these patients (in whom 39 lumbar levels were seen) there was a difference of 30% or more in the diameter of the dural sac seen on anteroposterior lateral myelograms during flexion and extension. These findings were compared with the dural sac compression shown on the CT myelograms for these patients. RESULTS In 10 of the 33 patients (12 levels), the CT myelogram underestimated spinal stenosis, as compared with the upright flexion-extension myelogram. In 5 levels, stenosis of 70% or more seen on flexion-extension myelography was measured as 50% or less on CT myelography. CONCLUSIONS In patients with positionally dependent spinal stenosis, CT myelograms may underestimate the severity of the spinal stenosis, compared with upright flexion-extension myelograms. While upright flexion-extension myelography is not a first-line imaging modality for patients with spinal stenosis, it should be considered for patients whose symptoms are not explained by routine cross-sectional imaging studies to exclude positionally dependent spinal stenosis.