Understanding the factors underlying establishment and spread of exotic species in order to predict invasion risks is a major goal in invasion biology. Many theoretical studies investigated the ecological and evolutionary components of these factors and their impact on the invasive process. Yet, hypothesis tests through experimental approaches are still scarce because of the practical and ethical difficulties associated with the introduction of exotic species in nature. Thus, most empirical results come from a posteriori analyses of fortuitous invasions, which allow correlative approaches at best and give no information about invasion failures. In this paper, we propose microcosms, i.e. reproducible controlled simplified environments, as an alternative to experimental introductions in natura. From a review of the literature, we discuss the distinctive features of microcosms to test theoretical predictions about invasion. Our analysis focuses on studies involving populations in transitory dynamics after a demographic bottleneck and/or subject to an adaptive challenge, two key characteristics of invasive processes. Despite their small number, these studies have been used successfully to explore the influences of various factors, mainly related to the introduction site characteristics (its abiotic conditions and their spatial and temporal heterogeneity), and to a lesser extent to the introduced individuals themselves (propagule pressure, genetic diversity and adaptations in the introduced population) or the invaded community. We argue that microcosms, as model systems, can be powerful tools to test theoretical hypotheses. They must however be used with care, as they do not account for the same complexity as natural systems. They are thus complementary to theoretical studies and field surveys, and contribute to reinforce the predictive value of invasion biology by linking theory and experimentation.