Aquatic habitats are important potential sites for gene transfer between indigenous bacteria and released genetically engineered microorganisms (GEMs). Legislation governing GEM release, and other practical considerations, have resulted in microcosms, of varying complexity, being used to study gene transfer in aquatic environments. This article reviews these microcosms, with particular emphasis on the more complex designs and, where possible, compares gene transfer results obtained in them with in situ studies. We conclude that microcosms can give results that are consistent with those obtained in situ and thus can be relied upon to give realistic predictions of in situ behaviour.