One hundred twenty-one Liberian children were admitted in coma to the ELWA Hospital, Monrovia, Liberia. Admitting diagnoses, before lumbar puncture, were compared with discharge diagnoses. Ninety-four children were discharged with a final diagnosis of cerebral malaria and 27 with a diagnosis of meningitis. The admitting diagnosis was correct in 76.6% (72 of 94) of patients with cerebral malaria and 59.3% (16 of 27) of patients with meningitis. The cerebrospinal fluid leukocyte count was the single most significant factor in determining the correct diagnosis. Without the cerebrospinal fluid analysis, the discriminant accuracy (77%), i.e. definitive separation of the two illnesses, was comparable to the physician's admission diagnosis (73%). Other data contributing to the differential diagnosis of cerebral malaria and meningitis included the number of days of fever before admission, the presence or absence of nuchal rigidity, fontanelle fullness and peripheral blood malaria smear. Mortality rates for cerebral malaria and meningitis were 14.9 and 29.6%, respectively. These data suggest that physicians cannot reliably discriminate between cerebral malaria and meningitis without cerebrospinal fluid analysis.