diagnosis of sleeping sickness (trypanosomiasis) is difficult because of the fluctuating levels of parasitaemia encountered in patients. in the present study we found that the polymerase chain reaction (pcr) demonstrated trypanosome infection in 20 out of 35 (57.1%) blood samples and in 21 out of 34 (61.7%) cerebrospinal fluid (csf) samples collected from an area endemic for sleeping sickness in north-west uganda. a total of 14 blood samples and 13 csf samples that were positive for trypanosomes by double centrifugation were also positive by pcr, demonstrating good concordance between the two methods. however, 6 (28.6%) of the 21 blood samples that were parasitologically negative were positive by pcr, while 8 (38.0%) out of 21 csf samples that were negative by double centrifugation were positive by pcr. these 14 negative samples could therefore be from sleeping sickness cases even though a positive pcr test is not evidence for the presence of trypanosomes. furthermore, of these 8 csf samples, 4 had been designated as early cases, based on the absence of trypanosomes and on a count of < 5 white blood cells (wbc) per ml. this suggests that some late-stage cases could potentially be missed according to the present criteria, and it is therefore important to perform clinical trials to determine whether these cases could be treated successfully with the first-stage drug alone. the remaining four csf samples had been classified as late-stage cases, based on a count of > 6 wbc per ml, even though trypanosomes could not be detected in these samples by either double centrifugation or pcr. a cut-off point of 5 wbc per ml, which is used as a rule of thumb to stage sleeping sickness patients, seems to leave some late-stage cases undetected since trypanosomes were detected in four csf samples from suspected cases with < 5 wbc per ml.