Though patient information leaflets (PILs) are provided to those invited to take part in medical research, they usually fall short in facilitating informed decisions about participation. We aimed to explore why there is an enduring requirement for a process that seems not to ‘work’, and to explain why the problems have proven resistant to correction. We analysed applications for ethical approval for 13 oncology trials and related official guidance. We interviewed 26 patients invited to participate in the trials. Data analysis was based on the constant comparative method. We show that PILs function latently to satisfy purposes other than their manifest function as a decision-facilitating tool. PILs are the outcome of a process of institutional scripting that is strongly shaped by the accountability demands inherent in the ethical review process. This results in the PIL being made to serve purposes both as a prospectus and as a contract. Though PILs have value for some patients, most do not recognise these documents as operating primarily in their interests. Patients make decisions in ways that deviate from official ideals. This analysis is important in recognising that no simple technical fix is available, and in enhancing sociological understanding of the institutional role of documents.