We tested the hypothesis that premedication with i.m. midazolam decreases core temperature dose-dependently. We studied six male volunteers, in random order, on 3 days: (1) no midazolam administration (control day), (2) midazolam 0.025 mg kg-1 i.m., (3) midazolam 0.075 mg kg-1 i.m. On the first day, subjects were maintained alert during a 30-min control period. On the second and third days, midazolam 0.025 or 0.075 mg kg-1 was administered i.m. Core temperatures were measured at the right tympanic membrane. Four adhesive skin surface probes were fixed on the chest, upper right arm, lateral calf and thigh. Finger tip perfusion was evaluated using forearm minus fingertip and calf minus toe, skin surface temperature gradients. Thirty minutes after midazolam i.m., the level of sedation in the volunteers was assessed. Peripheral venous blood was obtained immediately after the assessment of the level of sedation. Tympanic membrane temperatures after administration of midazolam 0.075 mg kg-1 i.m. were significantly lower than those on the control and midazolam 0.025 mg kg-1 i.m. days at 20 and 30 min. The decreases in tympanic membrane temperatures at 30 min after midazolam i.m. became larger as the volunteers were more deeply sedated. i.m. midazolam produced a concentration-dependent decrease in tympanic membrane temperature at 30 min after midazolam 0.025 and 0.075 mg kg-1 i.m. We conclude that midazolam impaired tonic thermoregulatory vasoconstriction, allowing core-to-peripheral heat redistribution in a dose-dependent manner after i.m. administration.