A comparison was made of the long-term impact of different methods of administration of chemotherapy (oxamniquine, 30 mg/kg in divided doses; or praziquantel, 40 mg/kg) on prevalence and intensity of Schistosoma mansoni infection in four areas in Kangundo Location, Machakos District, Kenya. In Area A, treatment was offered in October 1983 and again in April 1985 to all infected individuals. In Area H, treatment was offered in April 1985 to individuals excreting ≥ 100 eggs per gram (epg) of faeces. In Area H, treatment was offered in April 1985 to all infected school children, within the framework of the primary schools. In the witness area, Area W, treatment was given in April 1985, for ethical reasons, to a small number of individuals excreting ≥ 800 epg. Prevalence and intensities of infection were subsequently monitored at yearly intervals for three complete post-treatment years. In the Area S schools, clinical examination was also carried out at yearly intervals. Treatment of all infected individuals on two occasions (Area A) was the most effective and long-lasting way of reducing prevalence and intensity of infection. In this area, however, some earlier interventions had been carried out and pre-treatment intensities were lower than in the other areas. Treatment only of infected schoolchildren (Area S) also had a marked and prolonged effect, comparable to or better than treatment of individuals with heavy infections (Area H). Treatment of infected schoolchildren also caused a persistent reduction in the prevalence of hepatomegaly, and there was suggestive evidence from intensities of infection in community stool surveys (but not from incidence rates) of an effect on transmission. In all study areas, reinfection was most rapid and most intense among children. These findings are discussed in the light of theoretical considerations and of results from other studies, both on schistosomiasis and on intestinal helminths. We conclude that, in areas of low morbidity such as Kangundo, chemotherapy of schoolchildren only, at intervals of up to 3 years, is a satisfactory way of producing a long-term reduction in both intensity of infection and morbidity.