The Environmental Protection Agency is required to set a standard for lead-contaminated house dust, but whether dust lead loading (·g/ft2) or concentration (·g/g) is more predictive of children’s blood lead levels, which dust collection method should be used, and which surfaces should be sampled are unknown. Using a random sample of sequential births, we enrolled 205 urban children, 12 to 30 months of age, who had lived in the same house since at least 6 months of age. Samples of dust were obtained from predetermined surfaces in each child’s residence using a wipe method and two vacuum methods, the Baltimore repair and maintenance method (BRM) and the dust vacuum method (DVM). Other potential sources of environmental exposure also were analyzed for lead, including soil, water, and paint. In general, dust lead loading is more predictive of children’s blood lead levels than is dust lead concentration. Dust lead loading as measured with the BRM sampler explained more of the variation in children’s blood lead levels than did wipe loading and DVM loading (13.7, 10.1, and 5.9%, respectively, adjusted for other significant predictors). The partial correlation between BRM lead loading and children’s blood lead was significantly different than that for DVM lead loading, but it was not significantly different than that for wipe lead loading. Of the four surfaces measured, noncarpeted floors and interior window sills or wells were significantly associated with children’s blood lead levels in multiple regression models. These data indicate that dust lead loading is more predictive of children’s blood lead levels than is dust lead concentration and that, to determine if a housing unit is safe for children, noncarpeted floors and interior window sills or window wells should be measured using either the BRM or wipe sampling method.