Benzene is myelotoxic and causes leukemia in humans after extended periods of high exposure; however, leukemia risks in humans at low exposures are uncertain.  Benzene occurs in the work environment and in outdoor air, but mostly at concentrations below one ppm.  It is therefore important to assess the risk of benzene exposure to humans at low concentrations.  In this talk, we developed a physiologically based pharmacokinetic (PBPK) model for the uptake and elimination of benzene in humans to relate the concentration of inhaled benzene to the tissue doses of benzene and its key metabolites.  To apply the mathematical model to data in humans, the mathematical model must be integrated into a statistical framework that acknowledges the sources of variation in the data due to inherent intra- and inter-individual variation, measurement error, and other data collection issues.  The main contribution of this work is the estimation of population distribution of key PBPK model parameters using a fully parametric method.  In particular, we employ a Markov Chain Monte Carlo (MCMC) technique to fit the mathematical model to three sets of data producing samples from the posterior distributions of the parameters, from which inference on the parameters may be carried out.  We hypothesized that variability in metabolic parameters observed in earlier studies would be sufficient to explain observed variability in benzene pharmacokinetics.  The resultant simulations captured some but not all of the observed variability, indicating that one must also account for variability in physiological parameters, such as organ weights, to faithfully predict the full human population variability.  (This research was supported by the American Chemistry Council and CIIT.)