The performance of a full-scale granular activated carbon (GAC) treatment system in removing mutagens from drinking water obtained from the Ohio River has been evaluated using two bacterial mutagenicity tests. The Salmonella microsome assay (Ames Test) and a fluctuation assay were both performed using Salmonella typhimurium strains TA98 and TA100. Influent and effluent waters were collected at two GAC adsorbers, one filled with virgin GAC and one with nearly exhausted GAC. The samples were submitted to reverse osmosis (RO) pre-concentration, sequential liquid liquid extractions and XAD-2 resin adsorption. The RO aqueous concentrations of both influents gave positive mutagenic responses with both strains in the fluctuation assay but no activity in the Ames test. The extracts and adsorbates showed mutagenic responses in the Ames test with both strains, the highest values being observed with TA100 in the absence of metabolic activation. The summation of mutagenic activity on the basis of net revertants per liter indicated that exhausted GAC removed a substantial fraction (more than 85%) of the mutagenic activity whereas virtually complete removal was observed with virgin GAC. These data suggest that short-term mutagenicity tests may be useful in evaluating the performance of GAC or other adsorbents used in the treatment of drinking water.