This paper describes a prospective, double blind, randomised and dummy-controlled trial in 28 patients with chronic mechanical low back pain presenting to the York Pain Clinic. The therapeutic effects of epidural methyl prednisolone (80 mg) were compared with intrathecal midazolam (2 mg). All the patients had pain for a considerable length of time (range: 1-35 years) and all had received previous treatments which had failed. The two groups of patients were comparable in terms of pain duration, demography, extent of disability, anxiety and depression and pain locus of control. The pain was assessed before and for 2 months after treatment using the short form McGill Pain Questionnaire as well as visual analogue and verbal rating scales for sensory and affective components of their pain experience; patients also completed a pain diary. Both treatments caused a similar improvement in one-half to three-quarters of the patients for 2 months in patterns of activity and sleep as well as in the sensory and affective components of the pain. However, although the improvement in the two groups was similar, all the patients treated with steroid were either taking more or the same amount of self-administered analgesic medication after their treatment, whereas between one-third and one-half of the midazolam-treated patients took less medication during the 2 month follow-up period. We conclude that intrathecal midazolam is an effective treatment for chronic mechanical low back pain. The mechanism responsible for this effect is discussed.