This paper explores some of the issues raised by the finding in a randomized controlled trial (RCT) that general practitioners (GPs) taking part failed to use a computerized evidence-based guideline, nor did it have any impact on patient outcomes. GPs are expected to 'make an initial decision on every problem [patients] may present' and to address psychological and social problems in addition to biomedical ones. The computerized guideline imposed an external, largely biomedical, agenda that superseded the patient's. This disrupted the normal pattern of GP consultations and it was therefore ignored. Guidelines for any particular disease are effective if backed up by a detailed programme of education and audit. However, the large number of different conditions seen in general practice means that it is impractical to have such programmes for more than a small fraction of the clinical workload. The reductionist assumptions underlying the construction of evidence-based guidelines from systematic reviews lead to inflexible recommendations on the management of disease. Anthropologists and sociologists make an important distinction between scientifically defined diseases and the culturally constructed experience of illness. Because GPs deal with patients suffering illness that may or may not result from disease, disease-centred guidelines often conflict with their needs and wishes. The development of evidence-based medicine (EBM) was intended as a tool to help doctors make sense of evidence in the context of individual patients' problems. Few GPs are skilled in it, and it has been appropriated by powerful expert groups such as guidelines developers and the pharmaceutical industry. It is suggested that more understanding of EBM by GPs leads to better informed decision making by them and their patients.