The concept of model qualification, i.e., discovering the domain over which a validated model may be properly used, was illustrated with LATEBLIGHT, a mathematical model that simulates the effect of weather, host growth and resistance, and fungicide use on asexual development and growth of Phytophthora infestans on potato foliage. Late blight epidemics from Ecuador, Mexico, Israel, and the United States involving 13 potato cultivars (32 epidemics in total) were compared with model predictions using graphical and statistical tests. Fungicides were not applied in any of the epidemics. For the simulations, a host resistance level was assigned to each cultivar based on general categories reported by local investigators. For eight cultivars, the model predictions fit the observed data. For four cultivars, the model predictions overestimated disease, likely due to inaccurate estimates of host resistance. Model predictions were inconsistent for one cultivar and for one location. It was concluded that the domain of applicability of LATEBLIGHT can be extended from the range of conditions in Peru for which it has been previously validated to those observed in this study. A sensitivity analysis showed that, within the range of values observed empirically, LATEBLIGHT is more sensitive to changes in variables related to initial inoculum and to weather than to changes in variables relating to host resistance. © 2005 The American Phytopathological Society.