This paper argues for the inclusion of ethnography as a research methodology for understanding the effects of public health policy. To do this, the implementation of DOTS (Directly Observed Therapy, Short-course) – the World Health Organization (WHO) prescribed policy for the control of the infectious disease tuberculosis – is explored in the context of Nepal. A brief history of DOTS and its implementation in Nepal is outlined, and the way it has been represented by those within the Nepal Tuberculosis Programme (NTP) is described. This is followed by an outline of the research done in relation to this, and the ethnographic methods used. These ethnographic data are then interpreted and analysed in relation to two specific areas of concern. Firstly, the effects around the epidemiological uses of ‘cases’ is explored; it is argued that a tightening of the definitional categories so necessary for the programme to be stabilized for comparative purposes has profound material effects in marginalizing some from treatment. Secondly, the paper examines some of the implications and effects relating to the way that the ‘directly observed’ component was implemented. The discussion explores how current debate on DOTS has been played out in some medical journals. It argues for the importance of ethnography as a method for understanding certain questions that cannot be answered by particular, and increasingly dominant, research ideologies informed by randomized controlled trials. This raises important issues about the nature of ‘evidence’ in debates on the relationship of research to policy.