Examining the validity of exposure metrics used in air pollution epidemiologic models has been a key focus of recent exposure assessment studies. The objective of this work has been, largely, to determine what a given exposure metric represents and to quantify and reduce any potential errors resulting from using these metrics in lieu of true exposure measurements. The current manuscript summarizes the presentations of the co-authors from a recent EPA workshop, held in December 2006, dealing with the role and contributions of exposure assessment in addressing these issues. Results are presented from US and Canadian exposure and pollutant measurement studies as well as theoretical simulations to investigate what both particulate and gaseous pollutant concentrations represent and the potential errors resulting from their use in air pollution epidemiologic studies. Quantifying the association between ambient pollutant concentrations and corresponding personal exposures has led to the concept of defining attenuation factors, or alpha. Specifically, characterizing pollutant-specific estimates for alpha was shown to be useful in developing regression calibration methods involving PM epidemiologic risk estimates. For some gaseous pollutants such as NO2 and SO2, the associations between ambient concentrations and personal exposures were shown to be complex and still poorly understood. Results from recent panel studies suggest that ambient NO2 measurements may, in some locations, be serving as surrogates to traffic pollutants, including traffic-related PM2.5, hopanes, steranes, and oxidized nitrogen compounds (rather than NO2).