The relevance of the micronucleus test to human studies was investigated by using bone marrow from leukemic patients treated with antileukemic drugs. The median incidence of micronucleated erythrocytes and erythroblasts, respectively, increased from control values of 0.04 and 0.72% to 0.29 and 25.3% in leukemic cases; and the frequency of micronucleated erythroblasts was inevitably higher than the control value in cases that showed a higher frequency of micronucleated erythrocytes, but the reverse was not true. These results indicate that almost the same changes of micronucleus formation that are observed in the mouse micronucleus test are produced in human bone marrow by antileukemic drugs--mutagenic compounds--and, if the micronuclei were scored restrictively in erythroblasts, the application of the micronucleus test to human bone marrow would be reasonable.