Verbal autopsies are being used widely to describe the causes of mortality and to assess the effect of interventions against specific diseases in developing countries where many deaths occur at home. A verbal autopsy has been in use in the Upper River Division of The Gambia since 1988. In this paper we present the results of a validation study of this technique. One hundred and forty-one verbal autopsies were reviewed on two occasions by the same three physicians. In 38 (27%) of the cases, the first and subsequent diagnoses differed. In 94 children admitted to Basse Health Centre, the results of verbal autopsies were compared with the diagnoses made by a paediatrician--only 44 (47%) matched. The poor sensitivity and specificity of the verbal autopsy in this study may have been due to the confounding effect of malaria, which can be difficult to distinguish from other causes of death in this community.