The insurance hypothesis states that communities with greater numbers of species will be more stable than communities with fewer species. Various theoretical realisations of this hypothesis and some empirical evidence have been interpreted as general support for the hypothesis. Here, I suggest that very few studies can claim to accurately test the hypothesis. In particular, this is because few studies are sufficiently long in duration, and because sometimes community level variability is measured inappropriately. There are very few experiments (perhaps only six) that meet the assumptions of the theoretical models closely. Across these, around a half of all tests indicate support for the insurance hypothesis, and half show no support. This is probably too small a number of experiments to draw any solid conclusions and should encourage more experimental and observational tests. In particular, there have been very few investigations of how environmental variability interacts with diversity to determine community stability. Other directions for future research include replacing species richness with diversity of responses to environmental change, and developing a more mechanistic framework for understanding empirical results.