The alkaline comet assay is an established, sensitive method extensively used in biomonitoring studies. This method can be modified to measure a range of different types of DNA damage. However, considerable differences in the protocols used by different research groups affect the inter-laboratory comparisons of results. The aim of this study was to assess the inter-laboratory, intra-laboratory, sample and residual (unexplained) variations in DNA strand breaks and formamidopyrimidine DNA glycosylase (FPG)-sensitive sites measured by the comet assay by using a balanced Latin square design. Fourteen participating laboratories used their own comet assay protocols to measure the level of DNA strand breaks and FPG-sensitive sites in coded samples containing peripheral blood mononuclear cells (PBMC) and the level of DNA strand breaks in coded calibration curve samples (cells exposed to different doses of ionising radiation) on three different days of analysis. Eleven laboratories found dose-response relationships in the coded calibration curve samples on two or three days of analysis, whereas three laboratories had technical problems in their assay. In the coded calibration curve samples, the dose of ionising radiation, inter-laboratory variation, intra-laboratory variation and residual variation contributed to 60.9, 19.4, 0.1 and 19.5%, respectively, of the total variation. In the coded PBMC samples, the inter-laboratory variation explained the largest fraction of the overall variation of DNA strand breaks (79.2%) and the residual variation (19.9%) was much larger than the intra-laboratory (0.3%) and inter-subject (0.5%) variation. The same partitioning of the overall variation of FPG-sensitive sites in the PBMC samples indicated that the inter-laboratory variation was the strongest contributor (56.7%), whereas the residual (42.9%), intra-laboratory (0.2%) and inter-subject (0.3%) variations again contributed less to the overall variation. The results suggest that the variation in DNA damage, measured by comet assay, in PBMC from healthy subjects is assay variation rather than variation between subjects.