The effects of intraperitoneal (IP) and lumbar intrathecal (IT) midazolam (MID) on nociception was studied in 38 male albino rats using the noxious tale-flick and hot-plate tests. Four groups received IP 0.1, 1, and 10 mg/kg MID or an equal volume of its vehicle benzyl alcohol 0.1 mg in 1 ml saline, while the other three groups received IT 10 and 100 micrograms MID or 0.5 microgram benzyl alcohol in 5 microliter saline. The two higher doses of IP MID produced statistically significant decrease of tale-flick latencies (P less than 0.005 and 0.05 at 10 and 100 mg/kg MID, respectively). This hyperalgesic effect could be seen, although the animals appeared highly sedated with reduced motor activity and relatively unresponsive to non-noxious stimuli. In contrast, IT injections of 10 micrograms MID produced moderate but statistically significant prolongation of tail-flick latencies (P less than 0.05) without noticeable change in motor activity. This analgesic effect could not be observed with the higher dose of IT MID until an hour after its administration. The almost complete immobility and ataxia seen after the high doses of IP and IT MID (animals lying on their sides) precluded reliable hot plate testing in these animals. Apparently part of the high IT dose of MID diffused into the brain, as observed after high-dose IP administration. We therefore propose that the analgesic effect of midazolam stems from its action at the spinal level, while its sedative and hyperalgesic effects are a function of its supraspinal action.