CommentaryRamoni et al. 1 is the latest in a recent series of papers 2-4 reporting the results of surveys and interviews with researchers exploring their views about return of results. The debate about returning the results of genomics research has been heated, including in this journal, but we argue here that results of surveys of investigators have little value in policy formation about this issue. We need to ask better questions.Given that a primary justification for returning research results is to improve participants' outcomes by reducing disease risk, 5,6 ethics research undertaken to inform this debate must engage health policy. What matters for policy formation-and what we need evidence about-is what people actually do. Yet it is commonplace to observe a wide gap between what people say they will do or would like to do and their ultimate actions. The results here are no different. Ramoni et al. found that although two-thirds of the investigators they surveyed thought that return of research results was warranted under at least some circumstances, <10% had actually returned results, observations similar to those reported by others. The first Electronic Medical Records and Genomics (eMERGE) consortium, for example, returned no results, despite extensive discussion about which results to return. 7 The fact that there is such a gap between survey responses and actions raises a fundamental question about what these surveys and interviews, which typically pose hypothetical questions to researchers whose connections with research participants range from clinician scientists and primary data gatherers to downstream data users, actually reveal. The research may simply be accessing idealized expressions of how the researcher respondents would prefer to act, unconstrained by resources and the attitudes and actions of others. But even if one were to take these responses as probative of investigators' attitudes, these studies are not getting at what actually drives action, nor about the circumstances encountered in the conduct of research that lead to the disconnect between stated attitudes and outcomes. For example, although Ramoni et al. seek to identify which responses distinguished those researchers who report believing that results should be returned from those who do not, they do not answer the question of how salient any of these responses are or have been in determining whether or not results were returned. An understanding of what conditions-and to what extent attitudes and whose-drive action in the conduct of research would provide more useful evidence for debate. That is, we need an understanding of how the conduct and contexts of research mitigate for or against actions such as the return of results.Regardless of the relationship between opinion and action, the question still remains of what weight researchers' views should be given. The issue is both as broad as the value of surveys and responses to hypothetical questions and scenarios in producing evidence about how the world is and should be, and as n...