The development of ways to evaluate interventions that may have an impact on quality of life is a rapidly-developing area of research in clinical oncology, especially within the context of randomized controlled trials. We propose a role for assessments of preferences in such evaluations, including preference studies designed to assess attitudes toward the clinical acceptability of interventions, and preference trials designed to assess choice behaviour in relation to interventions. We suggest that such preference assessments represent a specific case of a more general issue: the need to develop an 'ethics of evidence', that is, standards for the creation, assessment and communication of evidence. We then outline a framework within which an 'ethics of evidence' might be developed, and suggest that the framework also may provide a useful model for the processes involved in the transfer of research results into clinical practice. As an illustration, we consider the problem of decision making in circumstances where the choice of therapy depends primarily on the patient's own preferences, as, for example, in the choice of mastectomy or breast-conserving treatment in early-stage breast cancer. The long-term goal is to develop criteria which might be used to foster shared rational decision making in such circumstances.