The 32P-postlabelling assay is widely used for detection of carcinogen-DNA adducts and other types of modified nucleotides in DNA. The principle of the method is the enzymatic digestion of DNA to nucleotides, 5'-labelling of these nucleotides with an isotopically labelled phosphate group, and the resolution and detection of the labelled products. Since the development of the original procedure in the early 1980s, many methods have been developed to increase the sensitivity of the method by selection of the modified nucleotides prior to labelling. In favourable circumstances, the method can achieve a level of detection as low as 1 modification in 10(10) nucleotides and requires relatively small quantities of DNA (less than 10 microg). It has been used to detect and characterise DNA adducts formed by numerous genotoxic carcinogens in bacterial and mammalian cells, in animals and, in some cases, in human tissues. Most classes of carcinogen have been subjected to 32P-postlabelling analysis, ranging from bulky and/or aromatic compounds to small and/or aliphatic compounds; it has also been used, with modifications, to detect apurinic sites in DNA, oxidative damage to DNA, UV-induced photodimers and, to a lesser extent, DNA damage caused by cytotoxic drugs. It has provided the first clear evidence for the DNA-damaging properties of several synthetic carcinogenic hormones. It has revealed the DNA-damaging potential of complex mixtures such as coal-tar and tobacco smoke. It has been used in human biomonitoring studies to detect DNA damage from occupational exposure to carcinogens, and also from environmental (i.e. non-occupational) exposures. It has also led to the discovery of the presence of numerous modifications in DNA arising from endogenous processes. The rapid expansion in the use of the assay has resulted in some divergence of procedures and there is a case to be made for the use of more standardised protocols, particularly where human exposure to carcinogens is being measured and where such results may be required for risk assessment. While the procedure is quantifiable, the efficiency of adduct labelling is, in many cases, not quantitative, and the lack of adduct standards has, in many cases, limited the interpretation of data to a demonstration of higher adduct levels in exposed groups compared with unexposed groups. Future developments are expected in automation, standardisation and, in combination with other analytical methods, elucidation of the structures of the many DNA lesions whose existence has been revealed by the 32P-postlabelling technique.