AIMS--To study the accuracy of routine laboratory diagnosis of malaria with the aim of improving accuracy in diagnosis in the future. METHODS--A comparative study was made of all blood films submitted to two laboratories in London providing a slide-diagnostic service for malaria. RESULTS--There were 17 Plasmodium ovale infections, and of these only five (29.4%) were correctly diagnosed by the submitting laboratory; whereas of 210 other single species infections, 162 (77.1%) were correctly diagnosed (chi 2 = 18.4, p < 0.0001). There were six patients with mixed infections; only one (16.7%) was correctly diagnosed, whereas of 227 single species infections, 167 (73.6%) were correctly diagnosed (p = 0.007, using Fisher's exact test). There was no significant association between the presence of technical faults or numerous platelets and incorrect diagnosis. CONCLUSIONS--Plasmodium ovale and mixed infections were diagnosed incorrectly significantly more often than other species. The presence of technical faults or numerous platelets had no significant effect on whether or not submitting laboratories correctly diagnosed malaria.