Food chain theory provides explicit predictions for equilibrium biomasses among trophic levels in food chains of different lengths. Empirical studies on freshwater benthic food chains have typically been performed on chains with up to three levels and in field experiments with limited spatial and temporal scale. Here we use a ‘’natural snapshot experiment” approach to study equilibrium biomass and abundance among trophic levels in natural ponds differing only with respect to fish assemblage structure. Forty-four ponds were surveyed for their density and biomass of fish, snails and periphyton. Ponds were divided into three categories based on fish assemblage: ponds with no fish (two trophic levels), ponds with molluscivorous fish (three trophic levels), ponds with molluscivorous fish (three trophic levels) and ponds that also had piscivorous fish (four trophic levels). Ponds without fish had a high density and biomass of snails and a low biomass of periphyton, whereas snails with molluscivorous fish. In the presence of piscivores, molluscivore populations consisted of low numbers of large individuals. Snail assemblages in piscivore ponds were characterised by relatively high densities of small-bodied detritivorous species and periphyton biomass was not significantly different from ponds with three trophic levels. Thus, predictions from classic food chain theory were upheld in ponds with up to three trophic levels. In ponds with four trophic levels, however, there was a decoupling of the trophic cascade at the piscivore-molluscivore level. Gape-limited piscivory, predation on snails by molluscivores that have reached an absolute size refuge from predation, and changes in food preferences of the dominant snails are suggested to explain the observed patterns.