Data on maternal morbidity make it possible to assess how many women are likely to need essential obstetric care, and permit the organization, monitoring and evaluation of safe motherhood programmes. In the present paper we propose operational definitions of severe maternal morbidity and report the frequency of such morbidity as revealed in a population-based survey of a cohort of 20,326 pregnant women in six West African countries. The methodology and questionnaires were the same in all areas. Each pregnant woman had four contacts with the obstetric survey team: at inclusion, between 32 and 36 weeks of amenorrhoea, during delivery and 60 days postpartum. Direct obstetric causes of severe morbidity were observed in 1215 women (6.17 cases per 100 live births). This ratio varied significantly between areas, from 3.01% in Bamako to 9.05% in Saint-Louis. The main direct causes of severe maternal morbidity were: haemorrhage (3.05 per 100 live births); obstructed labour (2.05 per 100), 23 cases of which involved uterine rupture (0.12 per 100); hypertensive disorders of pregnancy (0.64 per 100), 38 cases of which involved eclampsia (0.19 per 100); and sepsis (0.09 per 100). Other direct obstetric causes accounted for 12.2% of cases. Case fatality rates were very high for sepsis (33.3%), uterine rupture (30.4%) and eclampsia (18.4%); those for haemorrhage varied from 1.9% for antepartum or peripartum haemorrhage to 3.7% for abruptio placentae. Thus at least 3-9% of pregnant women required essential obstetric care. The high case fatality rates of several complications reflected a poor quality of obstetric care.