Risk characterization is the final step of the risk assessment process as practiced in the U.S. EPA. In risk characterization, the major scientific evidence and "bottom-line" results from the other components of the risk assessment process, hazard identification, dose-response assessment, and exposure assessment, are evaluated and integrated into an overall conclusion about the risks posed by a given situation. Risk characterization is also an iterative process; the results of a specific step may require re-evaluation or additional information to finalize the risk assessment process. Risks posed by atmospheric emissions are an example of an involuntary human health risk which typically receives a great deal of public attention. Characterization of the risks posed by atmospheric emissions typically requires the use of mathematical models to evaluate: 1) the environmental fate of emitted pollutants, 2) exposures to these pollutants, and 3) human dose-response. Integration of these models results in quantitative risk estimates. The confidence in a quantitative risk estimate is examined by evaluating uncertainty and variability within individual risk assessment components. Variability arises from the true heterogeneity in characteristics within a population or an event; on the other hand, uncertainty represents lack of knowledge about the true value used in a risk estimate. U.S. EPA's 1997 Mercury Study will illustrate some aspects of the risk characterization process as well as the uncertainty and variability encountered in the risk assessment process.