Beginning with several empirical papers in the late 1970's, there has been considerable research concerned with assessing the effectiveness of such attempted safety interventions as on-product warnings and safety signs. The focus of research on warnings has shifted from a debate on whether warnings work to systematic investigation of the factors that do or could influence safety-related product-user behavior. From the perspective of safety, the logical test of a warning must be reduction of the frequency and/or severity of accidents and injuries. A taxonomy of available research methods is described; strengths and problems associated with each method are discussed. Although research on topics related to warnings may legitimately address a wide variety of psychological issues, informed safety policymaking should rely primarily on well-controlled real-world studies. Within the restricted aim of making unambiguous contributions to generalizations that can inform safety policy, some methodological cautions are appropriate for both researchers and practitioners.