Plant functional traits are viewed as key to predicting important ecosystem and community properties across resource gradients within and among biogeographic regions. Vegetation dynamics and ecosystem processes, such as aboveground net primary productivity (ANPP), are increasingly being modeled as a function of the quantitative traits of species, which are used as proxies for photosynthetic rates and nutrient and water-use efficiency. These approaches rely on an assumption that a certain trait value consistently confers a specific function or response under given environmental conditions. Here, we provide a critical test of this idea and evaluate whether the functional traits that drive the well-known relationship between precipitation and ANPP differ between systems with distinct biogeographic histories and species assemblages. Specifically, we compared grasslands spanning a broad precipitation gradient (∼200-1,000 mm/y) in North America and South Africa that differ in the relative representation and abundance of grass phylogenetic lineages. We found no significant difference between the regions in the positive relationship between annual precipitation and ANPP, yet the trait values underlying this relationship differed dramatically. Our results challenge the trait-based approach to predicting ecosystem function by demonstrating that different combinations of functional traits can act to maximize ANPP in a given environmental setting. Further, we show the importance of incorporating biogeographic and phylogenetic history in predicting community and ecosystem properties using traits.