Articles E cologists have long debated what factors control the trophic (feeding) structure and function of ecosystems. This is more than just a matter of determining "who eats whom"; ecologists have pondered whether there are fundamental rules for determining (a) how many trophic levels an ecosystem can support, (b) how much primary production is consumed by herbivores, and (c) whether resources from the bottom of the food chain, or consumers from the top, control biomass, abundance, and species diversity in food webs. These questions are not only fundamental to ecology but essential for conservation and management. For example, the loss of a top predator in a food web that is largely controlled by top-down forces may drastically alter biodiversity and ecosystem function (e.g., nutrient cycling), whereas the same loss may have little effect in a resource-controlled (i.e., bottom-up) food web. To answer these questions, ecologists have expended an enormous effort to understand the relative importance of pre-dation or parasitism (and, to a lesser extent, mutualism) and competition for resources in trophic organization. Three basic models of control of trophic structure have emerged from this endeavor. The first of these, the energetic model of food webs, holds that energy supply (from the bottom of food webs), in concert with the relative efficiencies of consumers, limits the number of trophic levels and the relative biomass of each level in natural ecosystems (Lindeman 1942). The second model, commonly known as the "green world" hypothesis (Hairston et al. 1960), states that predators and parasites exert top-down control on herbivore populations. According to this model, herbivores do not generally compete with each other, and plant resources are not limiting because herbivore population densities remain low as a result of top-down control. The third model (Menge and Sutherland 1987) hypothesizes that the relative effects of predation on species diversity vary as a function of environmental stress (e.g., exposure, desiccation, extreme temperatures) and productivity. Specifically, the Menge-Sutherland model suggests that pre-dation should be more important at low and intermediate levels of stress, because high stress limits the abundance of predators more than it limits herbivore competitors. Competition for resources should be more important at high levels of stress (and low levels of productivity). Various modifications and elaborations of these three basic models of food webs and trophic structure have proliferated in the past several decades (Oksanen et al. 1981, Power 1992). Empirical tests of the food web models, and modifications thereof, have been conducted mostly in non-human-dominated ecosystems ranging from marine environments to freshwater lakes and streams, tundra, deserts, forests, and grasslands, each test often producing a different answer (Con-nell 1983, Schoener 1983, Sih et al. 1985). Empirical tests and the development of theory for food web dynamics have historically involved human-dominated ecosystems, such as agroecosystems (Rosenheim 1998), or recovering agricul-Stanley H. Human activities dramatically change the abundance, diversity, and composition of species. However, little is known about how the most intense human activity, urbanization, alters food webs and trophic structure in biological communities. Studies of the Phoenix area, situated amid the Sonoran Desert, reveal some surprising alterations in the control of trophic dynamics. Species composition is radically altered, and resource subsidies increase and stabilize productivity. Changes in productivity dampen seasonal and yearly fluctuations in species diversity, elevate abundances, and alter feeding behaviors of some key urban species. In urban systems-in contrast to the trophic systems in outlying deserts, which are dominated by limiting resources-predation by birds becomes the dominant force controlling arthropods on plants. Reduced predation risk elevates the abundance of urban birds and alters their foraging behavior such that they exert increased top-down effects on arthropods. Shifts in control of food web dynamics are probably common in urban ecosystems, and are influenced by complex human social processes and feedbacks.