Most mutagens require metabolic activation or can be metabolically inactivated. Lipophilic xenobiotics are typically metabolized by introduction and/or modification of a functional group and subsequent conjugation with a hydrophilic endogenous compound to allow excretion. Functional groups are most often introduced by the various cytochromes P-450. The risk that a reactive intermediate is generated is relatively high at this stage. However, many reactive intermediates may be efficiently inactivated by other enzymes, which modulate functional groups or introduce hydrophilic moieties. In other apparently rare cases, activation may also occur during these metabolic steps. Since bacteria and most cultured mammalian cells used as targets in short-term in vitro carcinogenicity tests are largely deficient in the cytochromes P-450, exogenous metabolizing systems (purified enzymes, crude subcellular preparations, intact cells or (in host-mediated tests) whole animals) are routinely used in such tests. Purified enzymes and crude subcellular preparations supplemented with individual or several cofactors are useful in elucidating the enzymatic control of mutagenic metabolites. As crude subcellular preparations are easy to prepare and to store, they are the metabolizing system most frequently used (supplemented with NADPH) in short-term tests. They favour the enzymes preferentially involved in activation. In contrast, intact cells (e.g. freshly isolated hepatocytes) and host animals additionally elicit the various enzyme activities preferentially involved in inactivation, so the results strongly depend on the kind of metabolizing system used. The choice of system plays a pivotal role in tests in bacteria, which are deficient not only in the 'preferentially activating enzymes' but also in the 'preferentially inactivating xenobiotic-metabolizing enzymes'. Mammalian cells in culture retain substantial activity of the latter enzymes. In our experience, results from in vitro mutagenicity tests in which adequate levels of activity of 'inactivating enzymes' are provided, either in the added metabolizing system or in the target cells, correlate better with carcinogenicity than tests in which the 'activating enzymes' are largely favoured.