The linked questions of what is a community and how does it change through space and time are some of the most long-lasting and vexing issues in ecology. In part, this reflects the different scales of time and space from which researchers draw their data, yet the issue is of great practical importance because it impinges on decision making in conservation biology. In offering the potential for linking differently scaled studies from both ecology and palaeoecology, ancient lakes are ideal areas in which to investigate this problem. High degrees of endemism and morphological specializations in the faunas of these lakes suggest that ecological interactions among such organisms are also likely to be complex, highly deterministic and stable, as a result of long periods of coevolved interactions. Data from long-term studies of fish ecology in Lake Tanganyika support this notion of highly evolved interactions and community stability, and have served as a general model for understanding ecological interactions in ancient lakes. However, the highly endemic and diverse ostracods of Lake Tanganyika do not conform to this model. Ostracod species associations are extremely variable in space, probably as a result of local colonization and extinction events within local population patches. Whereas individual species show affinities for particular environments, the assemblage observed in a given locality is highly unpredictable. Palaeoecological analyses of sediment cores also show species composition to have varied greatly over periods of hundreds of years. In some cases, however, this assemblage variability is accompanied by stability in total species richness. These results suggest that several models of community assembly may be relevant to ancient lake faunas, and that caution must be exercised in applying any particular model of community structure or stability to conservation biology and biodiversity management.