This study set out to test three hypotheses explaining the existence of closed canopied versus open ecosystems in the same landscape and on the same parent rocks. Our data lent no support to the first hypothesis that forests are ancient and grasslands are anthropogenically derived through deforestation. Carbon isotope data suggests the opposite; both open and closed vegetation had C4 isotope signals in deeper soils, indicating that grasslands are the ancient system. The second hypothesis is that closed forests require nutrient rich soils and are replaced by open vegetation on nutrient poor soils. Despite open vegetation frequently havinglower nutrient concentrations than closed vegetation in our study area, nutrient stocks to 1 m far exceeded threshold requirements to build a forest at all our sites. We suggest that vegetation patterns in our study area are a result of a third hypothesis, that forests and savannas are alternative stable states each subject to different ecological processes with positive feedbacks maintaining each state. Fire in open grassy vegetation is a key disturbance maintaining the boundaries between the two states. Frequent fires may deplete nutrients in the surface layers of the soil, thereby slowing tree growth and further maintaining the open state of the vegetation. However, our data on the vertical distribution of nutrients show strikingly similar nutrient patterns in the alternative biome states, with high nutrient concentrations in the open sites, suggesting that nutrient limitation is not an important feedback helping to maintain open ecosystems in our study area. Our results suggest that, rather than being drivers of vegetation patterns in our study area, soil nutrients are responding to dynamic vegetation changes caused by other factors.