Geophagy has been observed in nearly every long‐term study of folivorous primates. Yet despite frequent observations of this behavior, conclusive explanations for soil consumption remain ambiguous. This study tests the most frequently proposed hypotheses for geophagy using data collected on the geophagic behavior of the Milne‐Edwards’ sifaka (Propithecus edwardsi) living in two forests with varying levels of disturbance within Ranomafana National Park, Madagascar. Field data were collected from December 2002 to November 2003, during which time soil samples were collected for mineral analyses from 102 sites selected for geophagy and 42 control sites at which no geophagy was recorded. Results indicate that control soils differ significantly between the two study sites and that this difference is primarily attributable to varying levels of habitat disturbance. The frequency of soil consumption also does not vary significantly by sex or between logged and unlogged habitats. Soil consumption significantly correlated with fruit/seed consumption overall, but to a lesser degree in logged compared with unlogged sites. Clay minerals found in soils likely prevent absorption of dietary toxins in the gut and/or may mediate gut pH. This provides strong evidence for the protection hypothesis for geophagy, which may be especially important in areas where logging, or other forms of habitat disturbance, has been experienced. Abstract in French is available with online material.