This study describes a method for measuring the relative oral bioavailability (RBA) of arsenic (As) in soil and other soil-like media using young swine as the animal model. Groups of animals are exposed to site soil or sodium arsenate orally for 12 d. Forty-eight-hour urine samples were collected from each animal on d 6–7, 8–9, and 10–11 and were analyzed for total As. The urinary excretion fraction (UEF) for each group was estimated by plotting the mass of As excreted in urine by each animal as a function of the dose administered, and then fitting a linear model to the data using simultaneous weighted linear regression. The RBA of a test material is calculated as the ratio of the UEF value for the test material divided by the UEF of the reference material. Uncertainty around the RBA estimate is calculated using Fieller's theorem. Application of this method to a series of test soils indicates that RBA values for As can range from 18 to 52%. This wide variability supports the conclusion that there may be important differences in RBA between sites, and that use of a site-specific RBA value is likely to increase the accuracy of risk estimates for exposure to As in soil.