This thesis considers how policy ‘works’ as an instrument of governance. I report an ethnographic study of health care policy in the contested field of hospital planning. In this study I explored how a controversial policy to centralise hospital services was presented in national policy documents and implemented in one locality in England. I identified a shift in the framing of the policy, from one that presented centralisation as a means of improving access and making services more responsive to patients, to one of clinical necessity. In the new framing plans to close hospital services were presented as clinical decisions that were based on the evidence and necessary to ensure safety. I have interpreted this framing as a rhetorical strategy deployed at both national and local levels with the aim of realising change in the face of community resistance to closing local hospitals and a concomitant policy rhetoric emphasising the need to involve the public in decisions about how services are provided. Although the persuasive power of the framing was limited, a more insidious form of power was identified in the way the framing disguised the political nature of the issue by defining it as a clinical problem. The framing had the effect of restricting the extent to which alternative courses of action could be considered, and undermined public participation in decisions about the delivery and organisation of services. My thesis is that, beyond the often quite obvious use of rhetoric, power operates in policy practices in ways that are often difficult to see. Central to these processes are medical knowledge and expertise which serve to frame the debate, shaping how the problem is understood, which solutions are considered (and which are not) and who is included in decision making. To the extent that health services research is orientated to problems as defined by policy makers and service managers, and in our adoption of the scientific paradigm, we are implicated in these processes. The knowledge and analytic techniques of health services research constitute an indirect control technology, shaping how an issue is understood, creating subjectivities and assigning authority, constructing versions of health care, and influencing the practice of health care professionals. I conclude by proposing a reorientation from ‘science’ to ‘scholarship’. Unlike the scientific paradigm which abstracts policy issues from their social and cultural context, scholarship assumes that policy issues can only be understood in their relational settings. Scholarship supports public deliberation of policy issues and is of practical benefit to decision-makers, opening up policy options and illuminating the different perspectives on policy issues and thereby contributing to more creative, more acceptable, and more effective policies.