Research Ethics Committees (RECs) are increasingly institutionalised as a feature of research practice, but have remained strangely neglected by social scientists. In this paper, we argue that analysis of letters from RECs to researchers offers important insights into how RECs operate. We report a traditional content analysis and an ethnographic content analysis of 141 letters to researchers, together with an analysis of the organisational and institutional arrangements for RECs in the UK. We show that REC letters perform three important social functions. First, they define what is deemed by a REC to be ethical practice for any particular application, and confer authority on that definition. They do this actively, through comments on particular aspects of proposals, and passively, through silences about other aspects. Second, they provide an account of the work of the REC, and function as a form of institutional display. Third, they specify the nature of the relationship between the REC and the applicant, casting the applicant in a supplicant role and requiring forms of docility. Writing and reading REC letters require highly specific competences, and engage both parties in a Bourdieusian "game" that discourages challenges from researchers. The authority of RECs' decisions derives not from their appeal to the moral superiority of any ethical position, but through their place in the organisational structure and the social positioning of the parties to the process thus implied. Letters are the critical point at which RECs act on researchers and their projects.