Geophagy has been associated with iron deficiency and anaemia, but no causal relationship has been established. To clarify this, we conducted a two-by-two factorial randomised, controlled trial on the effect of iron and multimicronutrient supplementation on geophagy in Zambian schoolchildren in Lusaka, from February to December 2001. Of the 406 children, 212 (52.2%) were girls and the mean (range) age was 10.2 (7-15) years. Geophagy was reported by 302 (74.4%) and more often in girls than in boys (80.2 vs. 67.7%, P = 0.007). The mean (range) daily earth intake was 25.2 (1-200) g. Geophageous children had more often geophageous relatives than non-geophageous children (79.5 vs. 1.9%, P < 0.001). Geophageous children had lower serum ferritin (20.5 vs. 25.0 microg/l, P = 0.032) but not haemoglobin (Hb) (129.2 vs. 130.4 g/l, P = 0.59), than non-geophageous. Among those with Hb < 130 g/l, geophageous children had significantly higher prevalence (53.7 vs. 30.6%, P = 0.024) of Ascaris lumbricoides infection than non-geophageous. The prevalence of geophagy (74.4 to 51.6%) and the intake of earth (25.3 to 15.0 g/day) declined (P = 0.001 and P < 0.001, respectively) among the 220 (54.2%) children followed-up. In bivariate analysis, non-iron supplementation reduced the prevalence of geophagy more than iron supplementation did, but this was not confirmed in the multiple logistic regression analysis. Multimicronutrients had no effect on either geophagy prevalence or earth intake. Geophagy was prevalent and associated with iron deficiency, but iron supplementation had no effects on geophageous behaviour. Geophagy could be a copied behaviour and the association between geophagy and iron deficiency due to impaired iron absorption following earth eating.