A methodology is developed for the assessment of the risk due to drinking water containing trace levels of nine selected volatile organic compounds (VOCs). The nine VOCs are; benzene, carbon tetrachloride, 1,1-dichloroethylene, 1,2-dichloroethane, dichloromethane, perchloroethylene, trichloroethylene, 1,1,1-trichloroethane, and vinyl chloride. The risk assessment contains two major parts; viz, a hazard assessment and an exposure assessment. The hazard assessment uses animal bioassay data and extrapolates the dose-response curve using four models; logit, probit, multistage and Weibull. The hazard and exposure assessments are combined to estimate the population risk and provide a risk reduction analysis. This latter analysis estimates the cases averted for different control levels. An extensive analysis is included of the uncertainties in all parts of the risk assessment. It is concluded that the uncertainty in exposure assessment is less that an order of magnitude and that the uncertainty in hazard assessment is 4-6 orders of magnitude. The largest contribution, by far, to the uncertainty in the risk assessment is due to the choice of model to fit the bioassay data.