Konzo, an upper motor neuron disease, was diagnosed in 116 subjects in a community-based survey of 28,500 inhabitants in the Tarime District in Tanzania. All cases had a uniform clinical picture with isolated non-progressive paraparesis of abrupt onset identical to the findings in earlier studies of konzo . The fishing population along Lake Victoria was not affected but farming villages only 5 km from the shore had a maximum prevalence of 14 per 1000 inhabitants. New cases of konzo have occurred annually since 1979 but 62 % were affected during an epidemic at the end of a drought in 1985. The geographical and temporal distribution is compatible with the proposed dietary aetiology of cyanide exposure from insufficiently processed cassava roots as the main factor. Other toxiconutritional factors may also be involved. An aetiological role of HTLV-I has earlier been excluded and the exclusion of the fishing population makes an involvement of another type of infectious agent unlikely.