Population sizes of two ice nucleation-active strains of Pseudomonas syringae were compared on leaves in controlled environments and in the field to determine the ability of microcosm studies to predict plant habitat preferences in the field. The P. syringae strains investigated were the parental strains of recombinant deletion mutant strains deficient in ice nucleation activity that had been field tested for their ability to control plant frost injury. The population size of the P. syringae strains was measured after inoculation at three field locations on up to 40 of the same plant species that were studied in the growth chamber. There was seldom a significant relationship between the mean population size of a given P. syringae strain incubated under either wet or dry conditions in microcosms and the mean population size which could be recovered from the same species when inoculated in the field. Specifically, on some plant species, the population size recovered from leaves in the field was substantially greater than from that species in a controlled environment, while for other plant species field populations were significantly smaller than those observed under controlled conditions. Population sizes of inoculated P. syringae strains, however, were frequently highly positively correlated with the indigenous bacterial population size on the same plant species in the field, suggesting that the ability of a particular plant species to support introduced bacterial strains is correlated with its ability to support large bacterial populations or that indigenous bacteria enhance the survival of introduced strains. Microcosm studies therefore seem most effective at assessing possible differences between parental and recombinant strains under a given environmental regime but are limited in their ability to predict the specific population sizes or plant habitat preferences of bacteria on leaves under field conditions.