Consumption of inorganic arsenic in drinking water at high levels has been associated with chronic diseases. Research groups have estimated historic exposure using databases and models of arsenic in drinking water supplies, along with participant residential histories. Urinary arsenic species are an established biomarker of recent exposure; we compare arsenic concentrations in historically collected urine samples with predicted estimates of arsenic exposure. Using a cohort of 462 subjects with at least one urine sample collected from 1984-1992 and an arsenic exposure estimate through drinking water at the time of the urine sample, individual exposure estimates were compared with speciated urine arsenic (UAs) concentrations using correlation and multiple regression analyses. Urine inorganic arsenic (UIAs) concentrations (trivalent arsenic, pentavalent arsenic, monomethylarsonic acid, dimethylarsonic acid) were best predicted by residential water arsenic concentrations (R(2)=0.3688), compared with metrics including water consumption (R(2)=0.2038) or water concentrations at employment locations (R(2)=0.2331). UIAs concentrations showed similar correlation when stratified by whether the arsenic concentration was predicted or measured. Residential water arsenic concentrations, independent of water intake or water concentrations at places of employment, best explain the variability in UIAs concentrations, suggesting historical reconstruction of arsenic exposure that accounts for space-time variability and water concentrations may serve as a proxy for exposure.