In spite of the ubiquity of omnivory in nature, its consequences for population dynamics have received little attention from theoretical and experimental ecologists. Having three direct consumer—resource links, three indirect numerical links, and a potential for indirect effects mediated by size structure and/or behavioral flexibility, a three—species omnivory system may exhibit complex population dynamics. In 2 X 3 m field enclosures in the littoral zone of a lake, I manipulated the dominating native benthivorous fish (>2—yr—old perch, Perca fluviatilis, omnivorous top consumer) and the dominating native benthic invertebrate predators (Sialis lutaria, Megaloptera, and odonates, intermediate consumers) in three gradients: increasing densities of perch in the near absence of Sialis and odonates, and increasing densities of Sialis and odonates in either the presence or absence of perch. The densities of their common prey (mainly chironomids) were left unmanipulated. Macroinvertebrate abundance, biomass, and size structure, as well as gut contents of perch and Sialis were sampled monthly over a 3—mo summer period. In the absence of perch, the experimental gradient of Sialis densities remained unchanged over time. In the presence of perch, Sialis decreased by about half at high initial densities, but remained unchanged at low initial densities. Perch also had a negative effect on odonates. In the near absence of Sialis and odonates, perch had a strong, negative effect on chironomids. Compared to enclosures without perch, chironomid abundance was strongly reduced at the lowest perch density, but leveled off with further increases in perch density. Sialis and odonates did not affect chironomid abundance when perch was absent. In contrast, chironomid abundance was positively affected by Sialis and odonates when perch were present. The overall effect of perch on chironomid abundance in the presence of Sialis and odonates was negative. The combined predatory and competitive effects of perch on Sialis and odonates raise the issue how Sialis and odonates coexist with perch. The lack of effect of perch on Sialis and chironomids at low densities of these prey suggests that prey refuges contribute significantly to their persistence in natural systems. The indirect positive effect of Sialis and odonates on chironomids indicates density dependence in per capita interaction coefficients and is most likely to be explained by a behavioral response of perch or chironomids to Sialis and odonates. I suggest that the theoretical and empirical investigation of model systems (such as three—component omnivory systems), which are sufficiently simple to be analytically and experimentally tractable, but still display a richness of common indirect effects, will contribute to our understanding of the dynamics of more complex food webs. Their study will benefit from experimental manipulations of more than one population at more than two densities.