Data collected from 200 children admitted to a hospital on the Kenyan coast who met a broad definition of severe acute respiratory infection (ARI) indicated that simple clinical signs alone are unable absolutely to distinguish severe ARI and severe malaria. However, laboratory data showed that marked differences exist in the pathophysiology of unequivocal malaria and unequivocal ARI. Children in the former group had a higher mean oxygen saturation (97 vs. 94, P < 0.001), mean blood urea level (5.3 vs. 1.9 mmol/L, P < 0.001) and geometric mean lactate level (4.5 vs. 2.1 mmol/L, P < 0.001), and lower mean haemoglobin level (5.3 vs. 9.0 g/dL, P < 0.001) and base excess (-9.4 vs. -2.6, P < 0.001) than those in the latter group. Using these discriminatory variables it was estimated that up to 45% of children admitted with respiratory signs indicative of severe ARI probably had malaria as the primary diagnosis. Radiological examination supported this conclusion, indicating that pneumonia characterized by consolidation was uncommon in children with respiratory signs and a high malarial parasitaemia (> or = 10,000/microliters). There is no specific radiological sign of severe malaria. In practice, all children with respiratory signs warranting hospital admission in a malaria endemic area should be treated for both malaria and ARI unless blood film examination excludes malaria. In those with malaria and clinical evidence of acidosis, but no crackles, antibodies may be withheld while appropriate treatment for dehydration and anaemia is given. However, if clinical improvement is not rapid, antibiotics should be started.