Micronucleus (MN) expression in peripheral blood lymphocytes is well established as a standard method for monitoring chromosome damage in human populations. The first results of an analysis of pooled data from laboratories using the cytokinesis-block micronucleus (CBMN) assay and participating in the HUMN (HUman MicroNucleus project) international collaborative study are presented. The effects of laboratory protocol, scoring criteria, and host factors on baseline micronucleated binucleate cell (MNC) frequency are evaluated, and a reference range of "normal" values against which future studies may be compared is provided. Primary data from historical records were submitted by 25 laboratories distributed in 16 countries. This resulted in a database of nearly 7000 subjects. Potentially significant differences were present in the methods used by participating laboratories, such as in the type of culture medium, the concentration of cytochalasin-B, the percentage of fetal calf serum, and in the culture method. Differences in criteria for scoring micronuclei were also evident. The overall median MNC frequency in nonexposed (i.e., normal) subjects was 6.5 per thousand and the interquartile range was between 3 and 12 per thousand. An increase in MNC frequency with age was evident in all but two laboratories. The effect of gender, although not so evident in all databases, was also present, with females having a 19% higher level of MNC frequency (95% confidence interval: 14-24%). Statistical analyses were performed using random-effects models for correlated data. Our best model, which included exposure to genotoxic factors, host factors, methods, and scoring criteria, explained 75% of the total variance, with the largest contribution attributable to laboratory methods.