This paper presents an analysis of data on the incidence of leukemia, lymphoma and myeloma in the Life Span Study cohort of atomic bomb survivors during the period from late 1950 through the end of 1987 (93,696 survivors accounting for 2,778,000 person-years). These analyses add 9 additional years of follow-up for leukemia and 12 for myeloma to that in the last comprehensive reports on these diseases. This is the first analysis of the lymphoma incidence data in the cohort. Using both the Leukemia Registry and the Hiroshima and Nagasaki tumor registries, a total of 290 leukemia, 229 lymphoma and 73 myeloma cases were identified. The primary analyses were restricted to first primary tumors diagnosed among residents of the cities or surrounding areas with Dosimetry System 1986 dose estimates between 0 and 4 Gy kerma (231 leukemias, 208 lymphomas and 62 myelomas). Analyses focused on time-dependent models for the excess absolute risk. Separate analyses were carried out for acute lymphocytic leukemia (ALL), acute myelogenous leukemia (AML), chronic myelocytic leukemia (CML) and adult T-cell leukemia (ATL). There were few cases of chronic lymphocytic leukemia in this population. There was strong evidence of radiation-induced risks for all subtypes except ATL, and there were significant subtype differences with respect to the effects of age at exposure and sex and in the temporal pattern of risk. The AML dose-response function was nonlinear, whereas there was no evidence against linearity for the other subtypes. When averaged over the follow-up period, the excess absolute risk (EAR) estimates (in cases per 10(4) PY Sv) for the leukemia subtypes were 0.6, 1.1 and 0.9 for ALL, AML and CML, respectively. The corresponding estimated average excess relative risks at 1 Sv are 9.1, 3.3 and 6.2 respectively. There was some evidence of an increased risk of lymphoma in males (EAR = 0.6 cases per 10(4) PY Sv) but no evidence of any excess in females. There was no evidence of an excess risk for multiple myeloma in our standard analyses.