Virulence of entomopathogens is often measured at the individual level using a single host individual or a group of host individuals. To what extent these virulence assessments reflect the impact of an entomopathogen on their host in the field remains largely untested, however. A methodology was developed to induce epizootics of the cassava green mite fungal pathogen Neozygites tanajoae under controlled conditions to evaluate population-level virulence of two (one Beninese and one Brazilian) isolates of the entomopathogen--which had shown similar individual-level virulence but different field impacts. In unrepeated separate experiments we inoculated mite-infested potted cassava plants with either 50 or 25 live mites (high and low inoculum) previously exposed to spores of N. tanajoae and monitored the development of fungal infections for each isolate under the same conditions. Both isolates caused mite infections and an associated decline in host mite populations relative to the control (without fungus) in all experiments, but prevalence of the fungus varied with isolate and increased with inoculum density. Peak infection levels were 90% for the Beninese isolate and 36% for the Brazilian isolate at high inoculum density, and respectively 17% and 25% at low inoculum density. We also measured dispersal from inoculated plants and found that spore dispersal increased with host infection levels, independent of host densities, whereas mite dispersal varied between isolates. These results demonstrate that epizootiology of N. tanajoae can be studied under controlled conditions and suggest that virulence tests at the population level may help to better predict performance of fungal isolates than individual-level tests.