The proportion of exfoliated buccal mucosal cells with micronuclei gives the opportunity to assess sensitivity to gamma-radiation and genotoxic compounds and in addition to monitor the effectiveness of cancer intervention strategies. So far, results on counting micronuclei in various publications are difficult to compare because of differences in methods used, especially with regard to microscopical magnification used and number of cells counted. The aims of this study were (i) to define a protocol for counting micronuclei; (ii) to assess the feasibility of manually counting micronuclei; and (iii) the assessment of inter- and intra-patient variability of the number of micronuclei. We propose the definition of a strict protocol on counting micronuclei, with regard to cytological preparation, definition of micronuclei, instrumentation, sampling of cells in a cytological specimen and sample size. Such a strict protocol is a prerequisite for counting micronuclei in exfoliated cells to get a reproducible and sensitive indicator of exposure and for cancer risk. Although the inter- and intra-observer reproducibility of counting micronuclei per 1000 cells using such a protocol is well, we show that the variability among 10 assessments of micronuclei per 1000 cells taken sequentially from a sample size of 10,000 nuclei of the same specimen can be enormous (coefficients of variation varied in seven individuals studied between 42.1 and 102.9%). Based on the observed low frequencies varying from 1.2 to 5.2 micronuclei per 1000 cells and the variation found, we conclude that at least 10,000 exfoliated cells should be screened to monitor a significant reduction of 50% in the number of micronuclei (for a patient with an initial frequency in the micronuclei frequency range given). Since it takes approximately 7 h to evaluate this number of cells, it is also concluded that counting of micronuclei requires automation.