Dozens of weather and climate modeling centers around the world use their own numerical models of land processes. Formulations and behaviors of these land models differ in many ways, clouding cross-model assessments. In an ‘intercomparison’ project aimed at evaluation and of land models, we demonstrated the existence of gross measures of model behavior that characterize and explain differences across models (Koster and Milly, 1997, abstract, paper). We showed that the greatest contributor to inter-model spread was the presence or absence of stomatal resistance in the model, and we identified physically unrealistic runoff parameterizations in several models (Chen et al., 1997, abstract, paper). We demonstrated that the most significant differences across models could be attributed to physically erroneous aspects of formulations, poor physical assumptions, and coding errors. When adjusted for these factors, underlying spread of model behaviors is much smaller than previously recognized. These papers have induced corrections and improvements in many of the models studied.