The Erosion Productivity Impact Calculator (EPIC) model was validated using long-term data collected for two southwest Iowa watersheds in the Deep Loess Soil Region, which have been cropped in continuous corn Zea Mays L. under two different tillage systems (conventional tillage and ridge-till). The annual hydrologic balance was calibrated for both watersheds during 1988-94 by adjusting the runoff curve numbers and residue effects on soil evaporation. Model validation was performed for 1976-87, using both summary statistics (means or medians) and parametric and nonparametric statistical tests. The errors between the 12-year predicted and observed means or medians were less than 10 percent for nearly all of the hydrologic and environmental indicators, with the major exception of a nearly 44 percent overprediction of the N surface runoff loss for Watershed 2. The predicted N leaching rates, N losses in surface runoff, and sediment loss for the two watersheds clearly showed that EPIC was able to simulate the long-term impacts of tillage and residue cover on these processes. However, the results also revealed weaknesses in the model's ability to replicate year-to-year variability, with r2 values generally below 50 percent and relatively weak goodness-of-fit statistics for some processes. This was due in part to simulating the watersheds in a homogeneous manner, which ignored complexities such as slope variation. Overall, the results show that EPIC was able to replicate the long-term relative differences between the two tillage systems and that the model is a useful tool for simulating different tillage systems in the region.