The pharmacokinetic-pharmacodynamic model developed here characterizes the relationship between simulated plasma concentrations of thiopental and two dichotomous endpoints determined at induction of anesthesia: loss of voluntary motor power (clinical endpoint), and burst suppression of the electroencephalogram (EEG endpoint). The model incorporated data from two separate thiopental patient studies: a pharmacokinetic study with 21 males, and a pharmacodynamic study with 30 males. In the pharmacodynamic study, cumulative quantal dose-response curves for the clinical and EEG endpoints were developed from observations made during a constant-rate infusion of thiopental. Population mean parameters, derived from the bolus pharmacokinetic thiopental study, were used to simulate concentration-time data for the 150 mg.min-1 thiopental infusion rate used in the dose-response study. A single biophase model incorporating the two endpoints was generated, combining the pharmacokinetic and pharmacodynamic data from the two groups. Estimates of the mean effective thiopental concentrations affecting 50% of the population (EC50S) for the clinical and EEG endpoints were 11.3 and 33.9 micrograms.ml-1, respectively. The half-time for equilibration between arterial thiopental and the effect compartment was 2.6 min. These results are in reasonable agreement with previously described quantal concentration-response data, and with pharmacodynamic models developed for graded EEG responses. Simulation of bolus doses of thiopental with the new model provided ED50s for the clinical and EEG endpoints of 265 mg and 796 mg, respectively; the dose predicted to produce loss of voluntary motor power in 90% of an adult male population was 403 mg. A model combining population pharmacokinetics with cumulative dose-response relationships could prove useful in predicting dosage regimens for those drugs with responses that are categorical.