If epidemiological studies can define priorities for action, anthropological analyses are needed to clarify the conditions for the possibility of health problems. This article illustrates some of the ways in which public health and anthropological research may complement one another. Every year, 250,000 of the world's 200 million pregnant women die in Sub‐Saharan Africa. The medical causes of death are known and what should be done to avoid these unnecessary deaths is also known: quality caesareans, use of magnesium sulphate, hygiene during childbirth, tests and transfusion. So, concretely, the question is why sundry reforms fail or struggle for effective application. Drawing from a complex system of observation set up in four different services for 4‐month periods, this article aims to specify the qualitative variables that are behind the deaths of parturients.