Diagnosis of malaria in children is difficult without laboratory support because the symptoms and signs of malaria overlap with those of other febrile illnesses such as pneumonia. Nevertheless, in many parts of Africa diagnosis of malaria must be made without laboratory investigation. Therefore, a scoring system has been developed to assist peripheral health care workers in making this diagnosis. Four hundred and seven Gambian children aged 6 months to 9 years who presented to a rural clinic with fever or a recent history of fever were investigated. A diagnosis of malaria was made in 159 children who had a fever of 38 degrees C or more and malaria parasitaemia of 5000 parasites/microL or more. Symptoms and signs in children with malaria were compared with those in children with other febrile illnesses to identify features which predicted malaria. Symptoms and signs were incorporated into various logistic regression models to test which were best independent predictors of malaria and these regression models were used to construct simple scoring systems which predicted malaria. A nine terms model predicted clinical malaria with a sensitivity of 89% and a specificity of 61%, values comparable to those obtained by an experienced paediatrician without laboratory support. The ability of peripheral health care workers to diagnose malaria using this approach is now being investigated in a prospective study.