The aim of this thesis is to explore how institutional cultures and policies relating to patient autonomy influence physicians’ clinical and ethical approaches to do-notresuscitate (DNR) decision-making at the end of life. I explore this in the context of the evolution of the American medical profession that shifted medical decision-making power from paternalism to patient autonomy. Decades ago, the “doctor knows best” attitude prevailed. Since then, the pendulum has swung towards honouring patient autonomy. This thesis explores the implications of these changes and considers the tensions between the principle of beneficence and patient autonomy. I conducted 58 semi-structured in-depth interviews with internal medicine physicians, sampled by years of experience and medical subspecialty. These interviews were conducted at three academic medical centres in the United States (University of Washington, Johns Hopkins, Columbia New York Presbyterian) and one in the United Kingdom (Addenbrooke’s Hospital). It emerged during the research process that two of these four hospitals had policies that prioritised patient autonomy whereas the other two hospitals had policies that prioritised making decisions in the patient’s best interest. The main focus of this dissertation is on the United States, where three of the hospital sites are located. The UK serves as an additional site as there are no hospitals in the United States that permit unilateral physician decision-making in the way that the UK allowed at the time of the study. The focus on autonomy in American medicine today highlights the importance of freedom and choice to make one’s own decision. However, to truly honour patient autonomy, the patient or surrogate must have an adequate understanding of his situation and choices. Although the shift away from paternalism was a necessary positive step, I argue that the way in which autonomy is practiced reflects a reductionist notion of autonomy that disempowers rather than empowers patients, as patients are asked to make choices without the understanding or guidance necessary to make informed choices. Particularly at hospitals where autonomy is prioritised over other ethical principles such as best interest, trainee physicians equate autonomy with giving a menu of choices. They are uncomfortable giving a recommendation based on clinical knowledge as they worry that would be infringing upon patient autonomy. However, trainee physicians feel ethically compromised and experience significant moral distress when they are asked to provide therapies that they perceive to be futile. Despite that, doctors are still power brokers in the physician patient relationship and are able to manipulate conversations in other ways to sway patients towards a decision that they believe is in the patient’s best interest. Through a Habermasian approach, I explore how trainee communication practices of using purposefully graphic descriptions of resuscitation to discourage that choice are pathologic and disempowering. Physicians are fully inculcated in their respect for autonomy but unintentionally resort to strategic forms of communication because they feel constrained to resist recommendations in the name of autonomy. Rather than focusing on an ideology of autonomy, the American medical system needs to move towards practices that embrace a more nuanced and empathetic form of autonomy that fosters a more open form of communication that allows for cocreation of consensus between doctors and patients.