Most toxicology studies focus on effects of contaminants during exposure. This is disconcerting because subsequent survival may be affected. For instance, contaminant-induced mortality can be later ameliorated by reduced competition among the survivors, a concept we refer to as "density-mediated compensation." Alternatively, it can be exacerbated by toxicant effects that persist or appear after exposure, a phenomenon we term "carryover effects." We developed a laboratory framework for testing the contribution of exposure, density-mediated, and carryover effects to net survival, by exposing embryos and larvae of the streamside salamander (Ambystoma barbouri) to atrazine (0, 4, 40, 400 ppb; 3 ppb is the U.S. drinking water maximum) and quantifying survival during and 14 months after exposure. Atrazine is the most commonly used herbicide in the United States and a documented endocrine disruptor. We show that atrazine-induced mortality during exposure was ameliorated by density-dependent survival after exposure, but complete density-mediated compensation was precluded by significant carryover effects of atrazine. Consequently, salamanders exposed to >or=4 ppb of atrazine had significantly lower survival than did control animals 14 months postexposure. The greatest change in survival occurred at low exposure concentrations. These nonlinear, long-term, postexposure effects of atrazine have similarities to effects of early development exposure to other endocrine disruptors. Together with evidence of low levels of atrazine impairing amphibian gonadal development, the results here raise concerns about the role of atrazine in amphibian declines and highlight the importance of considering persistent, postexposure effects when evaluating the impact of xenobiotics on environmental health.