Abstract Following a dose of 4 Gy 250 kVp X-rays to the blood lymphocytes from three healthy donors the frequency of micronuclei was enumerated by three scorers. Eight slides were prepared from each blood sample with 1000 cytokinesis-blocked binucleate cells being scored for micronuclei and a recount undertaken 4 weeks later. Seven possible sources of variation were investigated: differences between donors, scorers, the four stages of sample preparation and time. A complete factorial analysis indicated that the major sources of variation were differences between donors and differences between scorers. One scorer identified fewer micronuclei than did the other two scorers in all slides of the irradiated blood samples and gave wide and inconsistent results for baseline counts. These results suggest that the performance of scorers should be screened before undertaking analysis of micronuclei. The most striking result, however, was the consistent performance of the scorers on the repeat counts compared with the very large differences between the scorers.