Data from 46 schools in western Kenya were used to investigate the performance of school-based questionnaires, on reported blood in stool and water-contact patterns, as indicators of the prevalence of human infection with Schistosoma mansoni. Prevalence of infection was associated with the prevalence of self-reported blood in stool, recent history of swimming and recent history of fishing. It was shown that use of a threshold of 30% of subjects reporting blood in stool would identify 42.9% of the 'high-prevalence' schools (i.e. prevalence > or = 50%) and 87.5% of the 'low-prevalence' schools (i.e. prevalence < 50%). A threshold of 25% reporting swimming would identify 57.1% and 93.7% of the high- and low-prevalence schools, respectively. Blood in stool appears to be too coarse an indicator to identify schools for mass treatment correctly. Although the use of multiple questions improved the diagnostic performance of the questionnaire in identifying the high-prevalence schools, it was unclear how questions can best be combined in other settings. However, there is a direct relationship between prevalence of S. mansoni infection and distance of the school from the lakeshore; analysis indicated that use of a threshold of 5 km from the lakeshore would correctly identify most (90%) of both the low- and high-prevalence schools. Distance to the lakeshore may therefore be used to screen schools in much of East Africa (i.e. those areas close to the Great Lakes where the infection is known to be prevalent and where much of the region's population is concentrated). In other areas of transmission, such as irrigation areas, further studies are still required.