We compared an rK39 dipstick rapid test (Amrad ICT, Australia) with a direct agglutination test (DAT) and splenic aspirate for the diagnosis of kala-azar in 77 patients. The study was carried out under field conditions in an endemic area of north-east Sudan. The sensitivity of the rK39 test compared with splenic aspiration was 92% (46/50), the specificity 59% (16/27), and the positive predictive value 81% (46/57). Compared with the diagnostic protocol used by Médecins sans Frontières, the sensitivity of the rK39 test was 93% (50/54), the specificity 70% (16/23), and the positive predictive value 88% (50/57). Compared with splenic aspirates, the sensitivity of a DAT with a titre > or =1:400 was 100% (50/50), but its specificity only 55% (15/27) and the positive predictive value was 80% (50/62). Using a DAT titre > or =1:6400, the sensitivity was 84% (42/50), the specificity 85% (23/27) and the positive predictive value 91% (42/46). All four patients with DAT titre > or =1:6400 but negative splenic aspirate were also rK39 positive; we consider these are probably 'true' cases of kala-azar, i.e. false negative aspirates, rather than false DAT and rK39 seropositives. There were no false negative DATs (DAT titre < or =1:400 and aspirate positive), but there were four false negative rK39 tests (rK39 negative and aspirate positive). The rK39 dipstick is a good screening test for kala-azar; but further development is required before it can replace the DAT as a diagnostic test in endemic areas of the Sudan.