We have studied the effects of midazolam premedication on multiple anaesthetic end-points (hypnotic, loss of verbal contact (LVC); motor, dropping an infusion flex or bag (DF); analgesic, loss of reaction to painful stimulation (LRP); and EEG, attainment of burst suppression (BUR)) during induction by slow thiopental infusion at a rate of 55 mg kg-1 h-1. Patients received midazolam 0.05 mg kg-1 i.v. (group TM, n = 12) or no midazolam (group T0, n = 13). ED50 and ED95 values and group medians for times and doses at the end-points were measured. Midazolam premedication reduced significantly thiopental ED50 and ED95 values at all end-points (exception for ED95 for BUR). Potentiation was greatest for the motor end-point (dropping the infusion bag (DF)) (ED95 +52%, ED50 +23%, median +39%), and smallest for painful stimulation (LRP) (median +18%; ED50 +13%). For LRP and DF, premedication was associated with significant, non-parallel increases in the slope of the thiopental dose-response curves, resulting in marked potency ratio changes from ED50 to ED95 (LRP +31%, DF +29%). There were no such increases for LVC or BUR. The interaction between midazolam and thiopental varied with the anaesthetic end-point and may also depend on the dose of thiopental. Our data suggest that the mechanism of interaction between midazolam premedication and thiopental was different for motor effects or analgesia (DF, LRP) compared with hypnotic effects or cortical depression (LVC, BUR), in agreement with the different central nervous system substrates underlying these distinct anaesthetic end-points.