Millions of people worldwide are exposed to drinking water containing arsenic, and epidemiologic studies have identified associations between the ingestion of arsenic-contaminated water and increased risks of cancer. In many of these studies, the assessment of arsenic exposure is based on a limited number of drinking water measurements, and the assessment of long-term or past exposure relies on the assumption that arsenic concentrations in sources of drinking water remain stable over time. In this investigation, the temporal stability of arsenic concentration was assessed in 759 wells in western Nevada state in the USA. Arsenic concentrations in these wells ranged from nondetectable to 6200 microg/L (median, 10 microg/L; standard deviation, 335 microg/L). Spearman correlation coefficients between arsenic concentrations measured in the same wells over a period of 1--5, 6--10, and 11--20 years apart were, respectively, 0.84 [95% confidence interval (CI), 0.81--0.86), 0.85 (95% CI, 0.81--0.88), and 0.94 (95% CI, 0.88--0.96). These findings suggest that, in this study area, arsenic concentrations in most wells remain stable over time and a limited number of measurements per well can be used to predict arsenic exposures over a period of many years.