It has been proposed that the best physical model of erosion from a plot is provided by a replicate plot (Nearing, 1998). Event data from paired bare fallow plots in the USLE database were used to examine the abilities of replicate plots, the USLE and the USLE-M to model event erosion on bare fallow plots. The Nash-Sutcliffe efficiency factor as applied to logarithmic transforms of the data was used to evaluate the overall performance of models at a number of locations. The value of this efficiency factor is influenced by both systematic and stochastic differences between the pairs. Systematic differences are the result of systematic differences in event runoff or event sediment concentration or both, and the degree of the impact of them varies as the regression coefficient for the relationship between the soil losses from the pairs varies from the value of 1.0. In most cases the replicate model performed better than the USLE-M that modelled event soil loss as a product of observed event runoff and event sediment concentration directly related to the EI30 index. Generally, failure of replicates to match runoff was compensated by the ability of the replicated to determine sediment concentrations better than the USLE-M.