This study was undertaken to evaluate the population pharmacokinetic behavior of ethanol from breath ethanol measures and to see if these results could be used to establish the drinking history of our drinking drivers. The population consisted of 55 self-identified light to heavy drinkers. All had been arrested at least once for driving under the influence of alcohol. Sixteen were women, and 39 were men. Breath was analyzed for ethanol using a 3-wavelength infrared spectrophotometer. An iterative 2-stage Bayesian (IT2B) parametric modeling program was used first to obtain gamma, a measure of the relative magnitude of the intraindividual variability. The nonparametric adaptive grid (NPAG) maximum likelihood program, using gamma, was then used to obtain the full nonparametric joint parameter density. A 2-compartment Michaelis-Menten model was evaluated. The 2-compartment model gave a gamma of 1.75. Thus, the standard deviation (SD) of the nonassay sources of intraindividual variability was 1.75 times the SD of the assay itself for the 2-compartment model. The NPAG program gave the following means, medians, modes, and standard deviations for the 2-compartment model: ka (h(-1)) = 6.43, 5.46, 2.93, 4.58; Vmax (g/h) = 12.09, 11.90, 13.03, 3.73; Km (g/L) = 0.1273, 0.1367, 0.1991, 0.0528; Vc (L) = 31.32, 29.30, 24.88, 10.52; kcp (h(-1)) = 4.38, 1.30, 1.12, 6.16; and kpc (h(-1)) = 9.11, 2.47, 0.89, 8.98. These drinking drivers had a rate of metabolism of ethanol that was between that of moderate drinkers and confirmed alcoholics. Properly collected breath ethanol measures can be useful in a therapeutic drug-monitoring situation to obtain quick, accurate, and reliable measures of a patient's ethanol concentrations.