Rainfall simulation experiments are widely used to study erosion and contaminant transport in overland flow. We investigated the use of two rainfall simulators designed to rain on 2-m-long (2-m2) and 10.7-m-long (32.6-m2) plots to estimate overland flow and phosphorus (P) transport in comparison with watershed-scale data. Simulated rainfall (75 mm h(-1)) generated more overland flow from 2-m-long (20 L m2) than from 10.7-m-long (10 L m2) plots established in grass, no-till corn (Zea mays L.), and recently tilled fields, because a relatively greater area of the smaller plots became saturated (>75% of area) during rainfall compared with large plots (<75% area). Although average concentrations of dissolved reactive phosphorus (DRP) in overland flow were greater from 2-m-long (0.50 mg L(-1)) than 10.7-m-long (0.35 mg L(-1)) plots, the relationship between DRP and Mehlich-3 soil P (as defined by regression slope) was similar for both plots and for published watershed data (0.0022 for grassed, 0.0036 for no-till, and 0.0112 for tilled sites). Conversely, sediment, particulate phosphorus (PP), and total phosphorus (TP) concentrations and selective transport of soil fines (<2 microm) were significantly lower from 2- than 10.7-m-long plots. However, slopes of the logarithmic regression between P enrichment ratio and sediment discharge were similar (0.281-0.301) for 2- and 10.7-m-long plots, and published watershed data. While concentrations and loads of P change with plot scales, processes governing DRP and PP transport in overland flow are consistent, supporting the limited use of small plots and rainfall simulators to assess the relationship between soil P and overland flow P as a function of soil type and management.