No perfect method exists for estimating historical soil loss from mediumsize watersheds. One approach to maximizing the accuracy for soil-loss estimates is to use a variety of different methods and compare their results for convergence. This paper compares four different methods to estimate soil loss from two medium-sized, fifthorder watersheds in southeastern Minnesota. The four comparative methods are the soil-survey and soil-truncation method (including a reevaluation of a Soil Conservation Service (SCS) study), gully-erosion estimates, reservoir-sedimentation measurements combined with estimated sediment-delivery ratios, and estimates derived from the Universal Soil Loss Equation (USLE). Although each method uses different data and procedures, their estimates of historical soil loss are comparable, excepting those for gully erosion. The estimates converged around the USLE estimate, ranging from 44 to 128% of the USLE. The soil survey and soil truncation method produced the lowest estimates, ranging from 44 to 79% of the USLE. The USLE estimate for one watershed fell in the middle of a range derived from the SCS study. The USLE computations for both watersheds ranged from 85 to 118% of the reservoir sedimentation estimate. Historical gully erosion ranged from only 3 to 15% of the different estimates of total soil loss, indicating that estimates of gully erosion alone are not good surrogates for total soil erosion. The convergence between the other watershed soil-loss methods lends a measure of credibility to the broad accuracy of the estimates.