A comparative study of three in vitro metabolising systems was performed in combination with Chinese hamster V79 cells, at which point mutation to 6-thioguanine resistance was scored. The three metabolising systems used were: (1) rat liver microsomal fraction (S9-mix); (2) feeder layer of primary embryonic golden hamster cells, according to Hubermann's system; (3) in vitro perfusion of rat liver according to the system of Beije et al. As model substances dimethylnitrosamine (DMN) and benzo[ a]pyrene (BP) was used. The liver perfusion was more efficient than S9-mix as an activating system of DMN, while the feeder layer of embryonic cells was unable to activate this compound. The activation of DMN with S9-mix was dependent on the presence of NADP. By exposing the target cells in the liver perfusion at different distances from the liver the biological half life of the active metabolite of DMN could be estimated to less than 5 s. With BP the three metabolising systems showed reversed results as compared with DMN — both the feeder layer cells and S9-mix activated BP, the feeder layer cells being most efficient. With liver perfusion, the perfusate itself was totally negative. Only the bile showed a week mutagenic effect. These results are in accordance with the notion that intact liver cells perform both an activation and a subsequent deactivation of BP. Because of the importance of hepatic bio-transformation in chemical mutagenesis and carcinogenesis it is emphasised that a liver perfusion system could be used in a testing protocol for genotoxic effects as a valuable tool in order to analyse the mechanism of action of mutagenic and carcinogenic compounds detected in other test systems, for instance bacterial/microsomal tests.