Most halogenated compounds showing genotoxicity do not themselves react with macromolecules but are transformed to reactive metabolites. In order to explain the widely different genotoxicities of halogenated ethylenes, the 'optimum stability' theory of epoxides has been developed recently. An epoxide must be reactive to alkylate DNA, but stable enough to reach the target from the place of its formation. Also, pharmacokinetic aspects are important in accounting for differences in genotoxicities of closely related compounds; examples are provided. Reactive metabolites may bind to proteins, to lipids and coenzymes, and to DNA, but binding to targets other than DNA is far more common than DNA alkylation. An example is provided by investigations on 2,2'-dichlorodiethyl ether from which chloroacetaldehyde is formed. In contrast to vinyl chloride, which generates chloroethylene oxide (chlorooxirane) as well as chloroacetaldehyde, this compound leads to extensive covalent protein binding of metabolites, but not to DNA binding.