This thesis examines the role of risk in framing confinement and spatial control in contemporary mental health policy. It argues that geographers’ focus on ‘post-asylum’ geographies has meant that the continued role of confinement in mental health has not been sufficiently examined. While deinstitutionalisation resulted in a change to the spatial configuration of care for people with mental illness it did not mean that confinement of people with mental illness ceased. Indeed, as the thesis shows, there has been a renewed emphasis on confinement in a number of jurisdictions. The thesis argues that the concept and language of risk has been used in mental health policy to support this renewed focus on confinement. The thesis a) provides evidence for the continued and/or resurgent significance of confinement, b) explains how this resurgence relates to the increasing framing of mental health policy by risk and c) explains what this means in relation to the spatiality of social control of people with mental illness. The methodological approach is shaped by an emphasis on understanding the historical context of the use of risk in public policy. There are two key aspects of the methodology. First, the importance of ‘the history of the present’; this is a concern to understand the historical conditions of existence upon which contemporary practices exist. Second, a discursive analysis of public policy broadly framed around critical discourse analysis. The continued role of confinement and how it relates to risk in contemporary mental health systems is examined in two case studies, one England and Wales, the other Queensland, Australia. The thesis demonstrates that there has been a renewed focus on confinement in these jurisdictions, driven by a policy concern to exert greater control over risk and uncertainty. It rejects the argument proposed by Castel (1991), and other governmentality theorists, including Deleuze (1992) and Rose (2002), that moves towards decentralised control have been shaped by governance through the abstract factors of risk. In both of the case studies risk was used largely in an individualised sense, in which risk was identified as being embodied in particular individuals, rather than in an aggregative sense. The thesis concludes that the concept of risk has primarily been used to support the continuation of spatially fixed modes of control over people with mental illness, rather than to facilitate decentred forms of control of people with mental illness.