Two common features of biological communities are (a) complex interactions among species, which make community dynamics sensitive to initial conditions, and (b) spatial heterogeneity, which fragments large—scale ecological systems into a mosaic of patches, hereafter termed a "metacommunity." This computer simulation study examines the effect of complex interaction on the global and local dynamics of metacommunities. Patches are physically identical and differ only in the initial proportion of species that colonize the patches. The random variation is then magnified by deterministic interactions that cause patches to follow different trajectories based on initial conditions. After a period of interaction, individuals from all patches join in global pool of dispersers that colonize a new "generation" of patches. Complex interactions can have at least two important effects on metacommunity dynamics. First the number of species coexisting in the metacommunity can greatly exceed the number of species coexisting in any single patch, despite the fact that the patches are physically identical, the species do not differ in colonization ability, and stochastic effects are absent after the colonization stage. Second, complex interactions provide a new source of variation upon which natural selection can operate at the patch level, providing a mechanism for the evolution of functionally organized communities.