The problem of user acceptance of knowledge-based systems is a current concern in medical informatics. User acceptance should increase when system-builders understand both the needs of potential users and the context in which a system will be used. Ethnography is one source of such understanding. This paper describes the contribution of ethnography (and an anthropological perspective) during the first year of a three-year interdisciplinary project to build a patient education system on migraine. Systematic fieldwork is producing extensive data on the information needs of migraineurs. These data call into question some of the assumptions on which the project was based. While it is not easy to rethink our assumptions and their implications for design, using ethnography has enabled us to undertake this process relatively early in the project at a time when redesign costs are low. This should greatly improve our chances of building a system that meets the needs of real users, thus avoiding the troublesome problem of user acceptance.