Cancer mortality for the period from October 1950 through May 1992 was analyzed in atomic bomb survivors exposed in utero. Risk estimates for this group were also compared to those for survivors who were less than 6 years old at the time of exposure. The cohorts studied include 807 in utero survivors and 5,545 persons exposed during childhood with all members of both groups having estimated doses of at least 0.01 Sv. The comparison group includes 10,453 persons with little (<0.01 Sv) or no exposure. Analyses were limited mainly to cancer deaths occurring between the ages of 17 and 46. Only 10 cancer deaths were observed among persons exposed in utero. However, there is a significant dose response with an estimate of excess relative risk per sievert (ERR/Sv) of 2.1 (90% confidence interval of 0.2 to 6.0). This estimate does not differ significantly from that for survivors exposed during the first 5 years of life. The cancer deaths among those exposed in utero involved leukemia (2), female-specific organs (3) and digestive organs (5). Nine deaths occurred in females, where the excess risk for all solid cancers has a 90% confidence interval on the ERR/Sv of 1.6 to 17. Significant risks were found for cancers of the digestive system [90% confidence interval (CI) on the ERR/Sv of 0.7 to 20] and for female-specific cancers (90% CI on the ERR/Sv of 0.7 to 42). These risks do not differ significantly from those seen in females exposed as children. There were no deaths from solid cancer in men exposed in utero. The ERR/Sv has an upper 95% confidence bound of 2.5 which does not differ from that for exposed children, where the upper 95% confidence bound is 1.5. The sexes differ even when female-specific cancers are excluded from the comparison. Although there were only two leukemia deaths among those exposed in utero, the leukemia death rate for this group is higher than that in the comparison group (P = 0.054) with an exposure effect that is about half the magnitude and not significantly different from that seen after childhood exposure (P = 0.103). However, there is no evidence of a dose response among those exposed in utero because no high-dose leukemia deaths were observed, a result that differs considerably from that for those exposed as children. There is a need for caution in the interpretation of these data. First, the number of cancer deaths is small; second, there is unexplained significant difference in the mortality from solid cancer between the sexes; and third, the excess of leukemia in those exposed in utero is not reflected in an increasing dose response.