Scientists often use models to reduce the complexity of intractable problems and to increase their understanding of the natural world. These models are recreations of the system of interest, but they incorporate only the parts relevant to the problem at hand. 1 When evaluating the ecological risk posed by the use and disposal of chemicals, the dominant model is the single species toxicity test. Much information about toxicity can be obtained from traditional toxicity tests through observing the response of individual species to the chemical. However, many parts relevant to the problem of predicting ecological risk are excluded from this model. Interactions between species and between species and physical features of their environment are key determinants of ecological outcome. Examples include disease resistance, predation rate, replacement of sensitive species by more resistant ones, changes in communities of organisms used for food, avoidance, biologically mediated changes in toxicant fate, or chemically mediated habitat destruction. These interactions cannot be assessed in tests using an individual species, but are the mechanics by which communities 540are shaped and reshaped by toxicant stress. However, these interactions can be included in models for studying ecological risk by using community-level toxicity tests.