SummaryAs part of a cross-sectional study among 156 primary school children (median age 13 years, range 10᎐18) in Western Kenya, geophagy was assessed through interviews. 114 (73.1%) of these children reported eating soil daily. Haemoglobin levels were determined in all 156 children and serum ferritin concentrations in 135. The mean haemoglobin (Hb) concentration was 12.7 g/dl, and median ferritin concentration 27.2 g/l. Both the proportion of anaemic (Hb Ͻ 11.0 g/dl) and of iron-depleted (ferritin Ͻ 12 g/l) children was significantly higher among the geophageous children than among the nongeophageous (9.6% vs. 0% anaemia; P ϭ 0.037; 18.4% vs. 5.4% iron depletion; P ϭ 0.046). Serum ferritin and haemoglobin concentrations were not correlated (r ϭ 0. 13 5; P ϭ 0. 100). Multiple regression analysis showed that geophagy, hookworm eggs per gram faeces and malaria parasite counts per l blood were independent predictors of serum ferritin, when controlling for other helminth infections, age and sex, and socio-economic and educational background of the children's families and family size (y ϭ 36.038᎐11.247(geophagy) ᎐ 0.010(hookworm epg) ϩ 0.001(malaria parasite counts); R 2 ϭ 0.17). Multiple regression analysis with haemoglobin as dependent variable and the same independent variables did not reveal any significant predictors. Analysis of the soil eaten by the children revealed a mean HCl-extractable iron content of 168.9 mg/kg (SD 44.9). Based on the data on the amounts eaten daily and this mean iron content, soil could provide on average 4.7 mg iron to a geophageous child (interquartile range 2.1᎐7.1 mg), which is equivalent to 32% of the Recommended Nutrient Intake (RNI) for girls (interquartile range 14᎐48%) or 42% of the RNI for boys (interquartile range 19᎐63%). Iron depletion and anaemia are associated with geophagy, but only serum ferritin concentrations were shown to be dependent upon geophagy in the regression model. From the cross-sectional data no inference about causality can be made. To clarify the possible causal relationships involved, longitudinal studies and iron-supplementation intervention studies are needed.