The induction of mutation to purine analogue resistance was assessed in Chinese hamster V79-4 cells exposed to gamma-radiation. After irradiation, the cells were grown in non-selective medium for different time intervals before respreading into medium containing 0.5-0.7 mug/ml thioguanine. In some experiments colonies arising in thioguanine-medium were counter-selected in medium containing the glutamine analogue azaserine, which distinguishes mutants with very little activity of the enzyme hypoxanthine-guanine phosphoribosyl transferase. Only these mutants were increased in frequency by radiation, the maximum measured frequencies occuring in cells respread after 2 days growth in non-selective medium. With longer intervals of post-irradiation growth in non-selective medium a fraction of the induced mutants was lost, and after large doses of radiation it is doubtful if the maximum frequency observed after 2 days post-irradiation growth represents the true induced frequency. The detection of freshly-induced mutants seemed to depend upon the dilution and decay of products formed from the genes prior to their mutation by radiation, since (a) selection of mutants in a higher concentration of thioguanine (2 mug/ml) increased the post-irradiation growth interval required to detect the maximum frequency of induced mutants, and (b) with increasing duration of post-irradiation growth in the non-selective medium, induced mutant cells were able progressively to overcome the growth-limiting effects of the analogue, to give large colonies when respread in thioguanine-medium. The radiosensitivities of 7 isolated mutant lines were indistinguishable from that of wild type cells, but the mutants were at a slight disadvantage when grown in competition with wild type cells. This disadvantage was consistent with the expected fitness of mutants relative to wild type cells calculated from estimates of the spontaneous mutation rate and the mean spontaneous mutation frequency. The induction of mutation to thioguanine resistance was non-linear with dose, yielding induced frequencies per rad of 5-10(-8) to 3-10(-7), but a plot of induced mutation frequency against log surviving fraction gave an approximately linear relationship. The same linear relationship holds for recently-published data on human and mouse cell cultures, so that all three mammalian cell types exhibit the same fixed probability of mutation induction relative to the extent of inactivation caused by ionising radiation.