The field of toxicology has traditionally assessed the risk of contaminants by using laboratory experiments and a range of pesticide concentrations that are held constant for short periods of time (1-4 days). From these experiments, one can estimate the concentration that causes no effect on survival. However, organisms in nature frequently experience multiple, applications of pesticides over time rather than a single constant concentration. In addition, organisms are embedded in ecological communities that can propagate indirect effects through a food web. Using outdoor mesocosms, we examined how low concentrations (10-250 microg/L) of a globally common insecticide (malathion) applied at various amounts, times, and frequencies affected aquatic communities containing zooplankton, phytoplankton, periphyton, and larval amphibians (reared at two densities) for 79 days. All application regimes caused a decline in zooplankton, which initiated a trophic cascade in which there was a bloom in phytoplankton and, in several treatments, a subsequent decline in the competing periphyton. The reduced periphyton had little effect on wood frogs (Rana sylvatica), which have a short time to metamorphosis. However, leopard frogs (Rana pipiens) have a longer time to metamorphosis, and they experienced large reductions in growth and development, which led to subsequent mortality as the environment dried. Hence, malathion (which rapidly breaks down) did not directly kill amphibians, but initiated a trophic cascade that indirectly resulted in substantial amphibian mortality. Importantly, repeated applications of the lowest concentration (a "press treatment" consisting of seven weekly applications of 10 microg/L) caused larger impacts on many of the response variables than single "pulse" applications that were 25 times as great in concentration. These results are not only important because malathion is the most commonly applied insecticide and is found in wetlands, but also because the mechanism underlying the trophic cascade is common to a wide range of insecticides, offering the possibility of general predictions for the way in which many insecticides impact aquatic communities and the populations of larval amphibians.