OBJECTIVES: We aimed to establish the pathological types of stroke in two incident populations in Tanzania, one rural and one urban, and to examine the clinical utility of the Siriraj and Allen scores in identifying stroke sub-types. DESIGN: This prospective community-based study identified cases as part of a stroke incidence study. Each patient underwent a full assessment including recording demographic information, taking a medical and drug history, and physical examination. A computed tomography (CT) head scan was used to classify strokes as resulting from a cerebral haemorrhage or ischaemia. The results were compared with the Siriraj and Allen scores, obtained from clinical findings. RESULTS: One hundred and thirty-two incident stroke cases were identified in the rural Hai demographic surveillance site (DSS) and 69 in the urban Dar-es-Salaam DSS; 63 patients with stroke due to ischaemia or cerebral haemorrhage from Hai and 17 from Dares- Salaam had a CT scan within 15 days of the stroke. Stroke was identified as due to ischaemia in 52 cases (82.5%) and to cerebral haemorrhage in 11 (17.5%) in Hai, and as due to ischaemia in 14 cases (82.4%) and to cerebral haemorrhage in 3 (17.6%) in Dar-es- Salaam. In both sites Siriraj and Allen scores were found to be of little value in predicting stroke sub-type. CONCLUSIONS: The ratio of ischaemic to haemorrhagic stroke is much higher in our cohort than previously reported in sub-Saharan Africa, and is closer to that in high-income countries.