Psychometric research on risk perception has frequently been invoked as evidence for a distinction between actual risk as measured by experts and perceived risk as experienced by laypersons. According to this view, perceived risk represents a distorted version of actual risk, shaped by the ignorance, prior beliefs, and subjective personal experiences of non-experts. The goal of risk perception research, it follows, is to illuminate the factors that account for deviations between actual and perceived risks. By contrast, qualitative social and political analyses of risk perception question the validity of the actual/perceived dichotomy and suggest that all perceptions of risk, whether lay or expert, represent partial or selective views of the things and situations that threaten us. Drawing on social studies of science and risk, as well as studies of quantitative risk assessment, this paper identifies three common models that link risk perception to regulatory policy -- labeled for convenience the realist , the constructivist , and the discursive . It calls attention to the ways in which the assumptions underlying each model have influenced risk-related research and decisionmaking in the United States.