Physical barriers, such as rivers and roads, constrain the movement of animals, usually by preventing access to adjacent habitats and impeding dispersal. Fences are artificial barriers that are commonly used as a conservation tool to intentionally restrict movements of animals to within protected reserves. However, the potential edge-effect of fences on the behaviour of animals within reserves is poorly understood. We examined the effect of fences on the movement patterns of African elephant (Loxodonta africana), an ecosystem modifier, in Pilanesberg National Park, South Africa. We used linear and non-linear models to determine the relationship between minimum distance from fence and seasonal daily net displacement of six GPS-collared female elephant. Elephant movement patterns were best explained by a piecewise regression that showed a strong negative relationship between minimum distance from fence and daily net displacement up to a ‘‘breakpoint” distance of 2551 m in the dry season and 3829 m in the wet season. The effect of the fence dissipated beyond this distance in both seasons. The increased tortuosity in movement patterns of elephant in the central area of the reserve suggested that they used this area more intensively for foraging compared to the peripheral area, as confirmed by differences in habitat selection. This occurs despite there being no difference in habitat composition between these areas. The decreased use of areas near the fence and more intensive foraging in the central areas constitute an important edge effect of fences. Since elephant are ecosystem engineers, such edge-effects could potentially cascade throughout the reserve, adversely altering ecologically processes, particularly in reserves with a high edge-to-area ratio.