Two different classifications of discograms have been used in a prospective study of 279 injected discs in 100 patients. The five-stage classification of Adams, Dolan and Hutton (1986) showed increased degeneration in the lower lumbar discs and more degenerative changes in men than in women. Exact reproduction of the patient's pain on injection was more common in fissured or ruptured discs than in less degenerate discs, with 81% sensitivity and 64% specificity of the discogram for pain. The additional information obtained by comparing computerised tomography (CT) with discograms was minimal. Discography was found to be useful in the evaluation of chronic low back pain in patients whose ordinary CT scans, myelograms and flexion-extension radiographs were normal. In spondylolysis and spondylolisthesis, discography can disclose whether fusion needs to be extended above the lytic level, and it may show if the pain in patients who have had posterolateral fusion is discogenic. Thus, discography gives information which is useful in deciding whether to operate on patients with chronic low back pain.