For those with sleeping sickness, the consequences of delayed diagnosis include poor prognosis at treatment and an increased risk of tsetse infection. Data on their socio-demographic and clinical characteristics, health-seeking behaviour and delays in presentation and diagnosis were collected from 119 diagnosed cases of rhodesiense sleeping sickness in eastern Uganda. The median total delay, from onset of the illness to diagnosis, was 60 days. The median service-provider delay (30 days) was markedly longer than the median patient delay (17 days). Each of these delays was, however, considerable and independently associated with patients presenting with late-stage sleepiness, giving odds ratios and (95% confidence intervals) of 7.29 (3.10-17.14) and 2.98 (1.38-6.43), respectively. A blood examination at the first visit was also associated with the service-provider delay (odds ratio = 0.45; 95% confidence interval = 0.22-0.95). Most of the patients (77.4%) had either been referred to the local sleeping-sickness hospital by other members of their community or presented at the hospital on their own initiative; few had been referred by other components of the local health system. The results are disappointing, not only in showing long delays in diagnosis (and therefore in treatment) but also in indicating that much of the delay is attributable to the service provider failing to diagnose sleeping sickness among symptomatic individuals.