The objective of this study was to evaluate the kinetics and dynamics of midazolam when administered by three different infusion schemes, using electroencephalography to measure pharmacodynamic effects. In a three-way crossover study, 8 volunteers received midazolam (0.1 mg/kg) by constant-rate intravenous infusion. The durations of midazolam infusions for the three trials were 1 minute, 1 hour, and 3 hours. Plasma midazolam concentrations and electroencephalographic (EEG) activity in the 13- to 30-Hz range were monitored for 24 hours. Based on separate analysis of each subject-trial, mean values for volume of distribution and distribution or elimination half-life did not significantly vary. Central compartment volume and clearance differed among the three midazolam infusion trials; however, the magnitude of change was small. EEG activity in the 13- to 30-Hz range significantly increased for all three midazolam infusion trials. Plots of midazolam plasma concentration versus pharmacodynamic EEG effect for the 1-hour and 3-hour infusion trials did not reveal evidence of either counterclockwise or clockwise hysteresis. Plots from the 1-minute infusion trial demonstrated counterclockwise hysteresis, consistent with an equilibration effect-site delay. This was incorporated into a kinetic-dynamic model in which hypothetical effect-site concentration was related to pharmacodynamic EEG effect via the sigmoid E(max) model. Analysis of all three infusion trials together yielded the following mean estimates: maximum EEG effect, 16.3% over baseline; 50% maximum effective concentration, 31 ng/mL; and an apparent rate constant for drug disappearance from the effect compartment which approached infinity. Despite the delay in effect onset during the 1-minute midazolam infusion, midazolam infusions in duration of up to 3 hours produce CNS sedation without evidence of tolerance.