My talk addresses how beta diversity maintains forest ecosystems. My model system is a tropical mountain, where soil weathering proceeds potentially rapidly but its speed varies with altitude, geology and topography. Consequently, nutrient availability (particularly P) varies greatly from site to site. I analyzed the foliar and litter nutrients of the major canopy-tree species in relation to net primary productivity and biomass in 12 tropical rain forests that were widely spaced from each other on different positions of the mountain with contrasting soil nutrient availability. Mean foliar N and P concentrations (weight basis) and N/P ratios varied across the 12 forests, and reflected soil nutrient availability. Leaf-litter N/P ratios were much wider (ranging from 50 to 225) than foliar N/P ratios were (40 to 100). Thus, nutrient-use efficient species in terms of productivity replaced inefficient species where that nutrient was critical. Theoretically, there is a trade-off between nutrient-use efficiency versus water-use efficiency. My analysis demonstrates that there is an array of tree species that differ in the foliar stoichiometry of N and P. Adaptive shifts (i.e. beta diversity) among species occur between nutrient-use and water-use efficiencies, and maintain forest productivity and biomass on spatial gradients on the mountain.