We assessed the reproducibility of selenium levels in toenails, comparing concentrations in two sets of specimens collected about a year apart, from 80 women (40 pre-menopausal and 40 post-menopausal) in the period October 1990 to February 1992. The women were participants in a prospective study on hormones and diet in relation to prediagnostic breast cancer (the ORDET study) conducted in northern Italy. Toenail selenium was determined by electrothermal atomic absorption spectrometry. The data were log-transformed as they were not normally distributed. To assess reproducibility Pearson correlation coefficients (r) for the two selenium determinations were calculated in pre- and post-menopausal women, according to smoking status and acetone treatment. A weighted kappa statistic (k) assessed inter-quintile agreement between the two sets of measures. Toenail selenium levels were highly reproducible (r = 0.57, p < 0.001), especially in pre-menopausal women (r = 0.66, p <0.001). Smoking lowered selenium levels (mean difference of 0.24 microg/g, p < 0.05 between smokers and non smokers) but did not significantly influence reproducibility. Acetone treatment to remove nail polish did not modify selenium levels but affected inter-quintile agreement, with moderate agreement (k = 0.58, p < 0.001) when acetone was used at both or neither samplings; and fair non significant agreement (k = 0.39, p = 0.06) when acetone was used at one sampling but not the other. As selenium levels in toenails are highly reproducible, notwithstanding variation in selenium levels in food and long-term changes in individuals' food choices, toenail selenium may be a useful biomarker of selenium exposure, particularly since toenail samples provide a measure of long-term exposure. However age, smoking status and acetone treatment are possible causes of misclassification.