A virulent fungal pathogen,Entomophaga maimaigaHumber, Shimazu&Soper, was discovered in gypsy moth,Lymantria dispar(L.), populations across northeastern North America in 1989. It had never before been reported from North America, but many reports have documented the importance of this natural enemy in Japanese gypsy moth populations. We conducted experiments to estimate parameters for several components of theE. maimaiga/L. disparsystem, and an object-oriented simulation model was used to evaluate the dynamics of this system. Studies conducted with a 1984 Japanese isolate ofE. maimaigaformed the basis for this model. Successful infection by conidia required a median time of 8.69 h at 25°C. Data for period of lethal infection demonstrated that this pathogen cannot develop at30°C, and that mortality is high across all instars. Transmission experiments using caged saplings in the field were used to estimate a value for the proportion of conidia that successfully reach hosts and cause infection. Experiments with the model demonstrated results consistent with the existence of a host density threshold below which the rapid increase in secondary infection characteristic of epizootics does not occur. Primary infections throughout the field season were important in producing an adequate level of secondary inoculum for development of epizootics. Instar-specific larval behaviors were hypothesized as resulting in differential instar exposure to primary inoculum; inclusion of behavior-specific coefficients resulted in disease phenology more similar to observed patterns in the field. The model was also successfully tested using weather data from sites and dates when epizootics caused byE. maimaigawere known either to occur or not to occur. Variability in weather conditions across a larger geographic area caused heterogeneity in development of epizootics.