It is usual in science studies to focus attention on either science policy or laboratory practices. This paper examines the mutual shaping of science and policy through the case of the Human Genome Mapping Project in the UK. It is argued that the evolution of policy was not merely a series of administrative choices. The strategies, boundaries and accounting methods of the programme, together with the organization and conception of research among grant-holders, can be shown to reflect the, often competing, demands from the worlds of science and administration. As policy evolved, the question of what science to support was translated into debates over the best way to do science, what scientific knowledge was for and what was to count as worthwhile knowledge.