Simple models, based on Lotka-Volterra types of interactions between predator and prey, predict that enrichment will have a destabilizing effect on populations and that equilibrium population densities will change at the top trophic level and every second level below. We experimentally tested these predictions in three aquatic food web configurations subjected to either high or low nutrient additions. The results were structured by viewing the systems as either food chains or webs and showed that trophic level biomass increased with enrichment, which contradicts food chain theory. However, within each trophic level, food web configuration affected the extent to which different functional groups responded to enrichment. By dividing trophic levels into functional groups, based on vulnerability to consumption, we were able to identify significant effects that were obscured when systems were viewed as food chains. The results support the prediction that invulnerable prey may stabilize trophic-level dynamics by replacing other, more vulnerable prey. Furthermore, the vulnerable prey, such as Daphnia and edible algae, responded as predicted by the paradox of enrichment hypothesis; that is, variability in population density increased with enrichment. Hence, by describing ecosystems as a matrix of food web interactions, and by recognizing the interplay between interspecific competition and predation, a more complete description of the ecosystem function was obtained compared to when species were placed into distinct trophic levels.