A cross-sectional study was conducted among 285 school children aged 5-18 years in Nyanza Province, Western Kenya, to determine the prevalence of geophagy and the types and amounts of soil eaten. Stool samples were taken from a subsample of 53 (19%) and their silica content determined to compare the results with the reported geophagy. Geophagy was practised by 73% of the children. The prevalence decreased with age for both sexes up to age 15, then remained stable for girls between 15 and 18 years but continued to decrease for boys in that age range. Most children ate soil from the surface of termitaria; others preferred the edges of paths and gullies, material from the wall of huts, and a chalk-like, soft stone commonly found in the area. The soil was eaten dry and was occasionally ground, but not processed in other ways. All but 4 of the children practising geophagy reported to eat soil at least once daily. The median amount reported eaten was 28 g daily, ranging from 8 to 108 g. The reported amount of soil eaten daily was significantly correlated to the results of the stool silica determinations. Using the median of 1% silica of faecal wet weight as a cut-off point to distinguish geophageous children from non-geophageous, the examination of a single stool sample had a sensitivity of 76% and a specificity of 80% to detect a geophageous child compared to the interview method. The cultural context of geophagy and its potential health impact in terms of infection and nutrition need to be further investigated, and it is suggested that more school and community-based studies on geophagy in different societies should be undertaken.