Epidemiologic studies conducted in the US have not previously detected an association between regional drinking water arsenic concentrations and corresponding cancer occurrence or mortality rates. To improve our estimation of cancer risk and arsenic exposure in the USA, we have investigated the reliability of several exposure markers. In the current study, we specifically evaluated the long-term reproducibility of tap water and toenail concentrations of arsenic, and the relation between water, toenail, and urinary measurement. Subjects included 99 controls in our case-control study on whom we requested a household tap water sample and toenail clipping three to five years apart. Additionally, participants were asked to provide a first morning void sample at the second interview. Tap water arsenic concentrations ranged from undetectable (<0.01 microg/L) to 66.6 microg/L. We found a significant correlation between both replicate water and toenail samples (intraclass correlation coefficient = 0.85, 95% confidence interval = 0.79-0.89 for water, and intraclass correlation coefficient = 0.60, 95% confidence interval = 0.48-0.70 for toenails). The inter-method correlations for water, urinary and toenail arsenic were all statistically significant (r = 0.35, p = 0.0024 for urine vs water; r = 0.33, p = 0.0016 for toenail vs water and r = 0.36, p = 0.0012 for urine vs toenails). Thus, we found both toenail and water measurements of arsenic reproducible over a three- to five-year period. Our data suggest that biologic markers may provide reliable estimates of internal dose of low level arsenic exposure that can be used to assess cancer risk.