In this work we have investigated the correlation existing between a short-term genotoxicity test (DNA repair in rat liver cells) and carcinogenicity in rodents. The work is in the framework of a line of thinking that considers as a possibility the utilization of the quantitative component of the information obtained from genotoxicity tests. In a preliminary report for 25 compounds belonging to different chemical classes, a correlation coefficient of 0.36 was found between carcinogenic potency in small rodents and potency in autoradiographic repair. This level of correlation is comparable with similar levels found for many other short-term tests: Ames test, alkaline DNA fragmentation in vivo, DNA adducts in vivo, morphological transformation in vitro and SCE induction in vivo. Obviously, since only 25 compounds were examined, assessment was rather uncertain, and the subdivision of the set into subsets for different chemical classes would have generated groups too small for a meaningful statistical analysis. With a much larger set (80 compounds) we hoped to be able to discriminate different predictivities for different chemical classes. This seems important because the test could be much more suitable for one given class than for another. Previous investigations with different short-term tests have shown that these differences can indeed exist and be very great. In this respect it is potentially very encouraging that the test considered here showed a fair correlation with carcinogenic potency for aromatic amines. Many other tests that we have examined so far have shown little or no predictivity for this important class of chemicals.