A microcomputer was programmed to accept data on the history and physical findings of patients, with low-back pain, suspected of having a herniated lumbar intervertebral disc, then suggest a likely diagnosis, with probability, and make suggestions for further management. Formal decision analytic techniques were used to test for the threshold of diagnostic likelihood that would make the expected value of laminotomy for excision of a herniated disc greater than the expected value of non-surgical management. The program is recursive, using its results to update its data base, and become more intelligent. In a blinded evaluation, an expert could not detect a significant difference between the output of the computer and the diagnoses and treatment plans of 10 clinicians.