To assess the correlation between micronucleus induction and human carcinogenicity, the rodent micronucleus assay was performed on known and potential human carcinogens in the 6th MMS/CSGMT collaborative study. Approximately 100 commercially available chemicals and chemical groups on which there was little or no micronucleus assay data were selected from IARC (International Agency for Research on Cancer) Groups 1 (human carcinogen), 2A (probable human carcinogen) and 2B (possible human carcinogen). As minimum requirements for the collaborative study, 5 male mice were treated by intraperitoneal injection or oral gavage once or twice with each chemical at three dose levels, and bone marrow and/or peripheral blood was analyzed. Five positives and 2 inconclusives out of 13 Group 1 chemicals, 7 positives and 5 inconclusives of 23 Group 2A chemicals, and 26 positives and 6 inconclusives of 67 Group 2B chemicals were found. Such low positive rates were not surprising because of a test chemical selection bias, and we excluded well-known micronucleus inducers. The overall evaluation of the rodent micronucleus assay was based on the present data combined with published data on the IARC carcinogens. After merging, the positive rates for Groups 1, 2A and 2B were 68.6, 54.5 and 45.6%, respectively. Structure-activity relationship analysis suggested that the micronucleus assay is more sensitive to the genetic toxicity of some classes of chemicals. Those to which it is sensitive consist of (1) aziridines and bis(2-chloroethyl) compounds; (2) alkyl sulfonate and sulfates; (3) acyl-type N-nitroso compounds; (4) hydrazines; (5) aminobiphenyl and benzidine derivatives; and (6) azo compounds. Those to which it is less sensitive consist of (1) dialkyl type N-nitroso compounds; (2) silica and metals and their compounds; (3) aromatic amines without other functional groups; (4) halogenated compounds; and (5) steroids and other hormones. After incorporation of structure-activity relationship information, the positive rates of the rodent micronucleus assay became 90.5, 65.2 and 60.0% for IARC Groups 1, 2A and 2B, respectively. Noteworthy was the tendency of the test to be more sensitive to those carcinogens with stronger evidence human carcinogenicity.