Rare, parasitic plants pose an interesting challenge to restoration practitioners. In addition to the problems associated with small population size, rare parasites may also be limited by their host requirements. We examined how the performance of a rare Pacific Northwest hemiparasite, Castilleja levisecta, was affected by the availability of different host combinations in the greenhouse and in the field. Castilleja levisecta individuals were grown with two individuals of the grass Festuca roemeri, two individuals of the aster Eriophyllum lanatum, one individual of each of these species (a “mixed” treatment), or without any host. We did not find support for the complimentary diet hypothesis, which predicts that parasites grown with multiple host species perform better than individuals grown alone or with a single host. In the greenhouse, C. levisecta individuals grown in the mixed treatment had greater stem growth than those planted with F. roemeri, but did not differ from E. lanatum or no-host treatments. In the field, vole activity had indirect effects on C. levisecta survival mediated through host species: vole tunneling and C. levisecta mortality were strongly associated with host treatments including E. lanatum. Vole tunneling and C. levisecta mortality were strongly associated with host treatments including E. lanatum. Field survival of no-host and F. roemeri treatments were significantly higher than those grown with E. lanatum. Our results emphasize the importance of basing conservation decisions on experimental research conducted under conditions similar to those of the intended application, as greenhouse results were a poor predictor of field performance. For restoration of endangered hemiparasitic plants, we recommend planting with hosts that are not attractive to herbivores.