African elephants Loxodonta africana (Blumenbach) may profoundly affect vegetation and associated animal bio-diversity in savannas (Conybeare 2004, Skarpe et al. 2004). Understanding the patterns of habitat use by elephants is crucial to predict their impacts on ecosystems (Ben-Shahar 1993, Nelleman et al. 2002), particularly now that many populations are recovering from past culling events or poaching outbreaks (Blanc et al. 2007). Surface water is one of the major constraints on elephant distribution (Chamaillé-Jammes et al. 2007, Stokke & du Toit 2002), and accordingly, elephant impacts are higher in the vicinity of water (Ben-Shahar 1993, de Beer et al. 2006). However, waterhole selection by elephant remains poorly understood. Weir (1972) showed in Hwange National Park (hereafter Hwange NP), Zimbabwe, that elephant numbers at waterholes over 24 h increased with the sodium concentration of water on nutrient-poor Kalahari sands. His work has become widely cited in elephant studies as it remains the only one, to the best of our knowledge, to have studied elephant use of waterholes in relation to the mineral concentration of water. Weir's work, however, took place when elephant densities in Hwange NP were low, likely below 0.5 elephants km−2 as estimated by aerial censuses (Williamson 1975). Since then, the elephant population has increased dramatically, particularly since the halt to culling operations in 1986 (Chamaillé-Jammes 2006, Cumming 1981). The present elephant density is much higher, estimated to be over 2 elephants km−2 (Chamaillé-Jammes et al. 2007, in press), and is one of the highest in the world (Blanc et al. 2007). Increased density may modify ecological constraints and affect the hierarchy of habitat selection processes (Morris 2003), and the extent to which water-nutrient selection still constrains elephant distribution at high population density – when their impact on savanna vegetation is the highest – remains unknown.