The Japanese atomic bomb survivor incidence data set and data on five other groups exposed to ionizing radiation in childhood are analysed and evidence found for a reduction in the radiation-induced relative risk of cancers other than leukaemia with increasing time since exposure. Overall, reductions of 5.7-6.1 per cent per year of time since exposure are indicated, depending on the time at which the reduction is presumed to start, and all the reductions are statistically significant at the 5 per cent level. There is no significant heterogeneity in the speed of the reductions in relative risk with time by cohort, by cancer type, sex, or age at exposure group. There is a significant reduction of relative risk with increasing age at exposure, but adjustment for age at exposure does not markedly affect the time trends of relative risk. For all of the groups considered, there is a statistically significant increase in the excess absolute risk with increasing time since exposure. However, by contrast with the relative homogeneity of the time trends of relative risk, there is statistically significant heterogeneity by cancer type within the Japanese cohort (P = 0.05) and between the cohorts (P < 0.0001) in the speed of increase of the excess absolute risk with time since exposure.