Results of traditional laboratory bioassays may not accurately represent ecological (field) host specificity of entomopathogens but, if carefully interpreted, may be used to predict the ecological host specificity of pathogens being considered for release as classical biological control agents. We conducted laboratory studies designed to evaluate the physiological host specificity of microsporidia, which are common protozoan pathogens of insects. In these studies, 49 nontarget lepidopteran species indigenous to North America were fed five biotypes of microsporidia that occur in European populations of Lymantria dispar but are not found in North American populations of L. dispar. These microsporidia, Microsporidium sp. from Portugal, Microsporidium sp. from Romania, Microsporidium sp. from Slovakia, Nosema lymantriae, and Endoreticulatus sp. from Portugal, are candidates for release as classical biological control agents into L. dispar populations in the United States. The microsporidia produced a variety of responses in the nontarget hosts and, based on these responses, the nontarget hosts were placed in the following categories: (1) no infection (refractory), (2) atypical infections, and (3) heavy infections. Endoreticulatus sp. produced patent, host-like infections in nearly two-thirds of the nontarget hosts to which it was fed. Such generalist species should not be recommended for release. Infections comparable to those produced in L. dispar were produced in 2% of the nontarget hosts fed Microsporidium sp. from Portugal, 19% of nontarget hosts fed Microsporidium sp. from Romania, 13% fed spores of Microsporidium sp. from Slovakia, and 11% of nontarget species fed N. lymantriae. The remaining nontarget species developed infections that, despite production of mature spores, were not typical of infection in L. dispar. We believe it is very unlikely that these atypical infections would be horizontally transmitted within nontarget insect populations in the United States.