Extrapolating the effects of toxicants with either the fixed application factor approach or one of the species sensitivity distribution models currently in widespread use presumes that toxicant effects on single, individual-level endpoints (e.g., survival, fecundity, and growth) reflect effects at the population level. Here, we consider if extrapolations derived on the basis of individuallevel endpoints might be misleading with regard to risk assessment and, hence, risk management decisions for ecosystems. Both analytically and by simulation, we demonstrate that for populations with multiplication rates close to one, effects of toxicants at the population level likely are less than or equal to effects on individual life-cycle traits, suggesting that risk assessments based on the latter likely are protective of population-level impacts. We used Monte Carlo simulations to explore how both the frequency of different life-cycle types in a community as well as their relative sensitivity to toxicants influence the toxicant sensitivity distribution of the community as a whole. We compared the output of our simulations with predicted no-effect concentrations derived by an application factor approach and a log-normal distribution-based model, using no-observed-effect concentrations for juvenile survival as input variables in both cases. Our analyses suggest that current extrapolation approaches appear to be protective, and may often be very overprotective, but we have identified conditions in which this may not be the case. We recommend that additional consideration be given to the relative frequency of different life-cycle types, to the proportions of sensitive and insensitive taxonomic groups in communities, and to the role of density-dependent influences on population dynamics to ensure that we develop environmental quality criteria that are sufficiently—but not overly—protective.