Sediments from four inshore industrial sites and a reference site in the Great Lakes were extracted with solvents and characterized chemically for polycyclic aromatic compounds (PACs). An aqueous phase and a crude organic extract were obtained. The crude organic extract was further resolved into fractions A-2 (polycyclic aromatic hydrocarbons) and A-3 (nitrogen-containing polycyclic aromatic compounds), which were analyzed for PACs by gas chromatography and gas chromatography-mass spectrometry. The extracts and fractions were tested for mutagenicity in three assays: Ames, rat hepatocyte unscheduled DNA synthesis, and Chinese hamster ovary hypoxanthine-guanine phosphoribosyl transferase (CHO/HGPRT). Sediments from the industrial sites contained 27 to 363 μg/g total PACs; the reference site, less than 1 μg/g. Qualitative differences in the residue profiles among the sites were attributable to the probable sources of the PACs (petroleum versus combustion). Only one industrial site yielded measurable (0.1 μg/g or more) concentrations of individual nitrogen-containing PACs. In the Ames assay, only the highest doses of the A-2 fractions from two sites approached positive results. Conversely, the crude organic extract and A-2 and A-3 fractions from all sites induced unscheduled DNA synthesis. Crude organic extracts and the A-2 and A-3 fractions from all industrial sites gave well-defined dose-response relations in the CHO/HGPRT assay. We established the presence of chemical mutagens in sediment that could be correlated with neoplasms in fish from many of the sites; however, the mutagenicity of the sediment extracts was not completely related to the degree of contamination by PACs. We also discuss the utility of mutagenicity assays in the evaluation of complex chemical mixtures and recommend the use of a CHO/HGPRT-type assay in which cells are not required to proliferate in the presence of potential interfering chemicals.