The process of human health risk assessment (HRA) is often judged by its ability to predict adverse outcomes of particular environmental contaminants or exposures for individual humans. Likewise, environmental scientists often examine the adverse outcomes of chemical or physical hazards on individual species. This ecotoxicological approach to ecological risk assessment (ERA) fails to encompass the potential range of adverse outcomes to populations, communities, and ecosystems. Moreover, whereas the success of HRA can be evaluated by examining the health of individual humans, success of ERA cannot because (1) populations of species are the important unit ecologically, rather than individuals, (2) the overall structure and complexity of the system is important rather than the structure or organization within one species (i.e., humans in the case of HRA), and (3) the overall functioning of the system is important rather than only the functioning of one species. I suggest that the risk assessment paradigm that includes hazard identification, dose-response analysis, exposure assessment, and risk characterization should have a parallel phase or discipline of research: evaluation of predicted and actual outcomes. This phase, termed "predictive accuracy" in this article, is particularly critical for Ecological Risk Assessment because the potential outcomes may occur long after the initial perturbation.