Abstract Wild-type Autographa californica nucleopolyhedrovirus (AcNPV or AcNPV.WT), AcNPV expressing a scorpion toxin (AcNPV.AaIT), and AcNPV expressing a mutated juvenile hormone esterase (AcJHE.SG) were compared in their capability to produce epizootics in larvae of Trichoplusia ni infesting collards in a greenhouse microcosm. Larvae treated in four different ways were released into 1.8-m 2 microplots in week 1. The four treatments included (1) uninfected larvae (control), (2) 100% AcNPV.WT-infected larvae (WT), (3) 100% AcNPV.AaIT-infected larvae (AaIT), and (4) 1:1 ratio of AcNPV.WT-infected and AcNPV.AaIT-infected larvae (WT+AaIT). On a weekly basis, larvae were sampled and new, uninfected larvae were added to all plots. Sampled larvae were reared until death and then subjected individually to DNA–DNA dot-blot hybridization assay to determine the proportion of insects infected with each virus in each plot. The entire experiment was repeated with AcJHE.SG in the place of AcNPV.AaIT. Epizootics of AcNPV.WT lasted 8 weeks after a single viral release in the replicated greenhouse microplots. AcJHE.SG epizootics also lasted 8 weeks after viral release, but this virus and AcNPV.AaIT were both out-competed by AcNPV.WT. AcNPV.AaIT was no longer detected in the T. ni population by the fourth week after release. AcNPV.WT also increased to greater numbers in soil than AcNPV.AaIT or AcJHE.SG after 8 weeks. Thus, it was possible to induce 8-week epizootics of AcNPV.WT in replicated microplots under artificial greenhouse conditions, and the wild-type virus out-competed the recombinant virus for a niche in this greenhouse microcosm, which reduces the probability that the recombinant virus will persist in an agroecosystem.