Some have questioned the importance of the small effect sizes generally reported in epidemiological studies of neurotoxicity. To some extent, this reflects a failure to appreciate the critical distinction between individual and population risk. In the first part of the paper, arguments are marshaled to support the contention that small shifts in the mean value of a health index within a study sample can, under some circumstances, carry substantial implications for the health status of the population from which the study sample was drawn. Under such circumstances, a population-based approach to prevention might be as effective as a patient-based approach. The second part clarifies conditions under which this will be true (e.g., a monotonic relationship between a health index and risk of disease) and conditions under which it might not (e.g., a J-shaped relationship). In the third part, the relative levels of uncertainty in characterizing individual versus population risk are explored. In neurotoxicological studies, uncertainty in characterizing individual risk could be reduced by adjusting, in addition to bone fide confounders, for covariates that are strong predictors of outcome and by more assiduous efforts to characterize major effect modifiers.