These are exciting times for savanna ecologists. The past decade has seen a number of advances toward a general model to explain tree-grass coexistence in savannas. The search for a single paradigm for tree-grass coexistence is more than a century old, and was often deemed impossible because of the wide geographic, edaphic, and climatic range within which savannas are encountered. Despite the fact that savannas are characterized by complex and apparently destabilizing interactions between trees and grasses, tree-grass co-dominance persists across a wide range of environmental conditions. This begs the following question: do savannas emerge as the result of disparate processes acting under different conditions, or can a single theory explain their existence and distribution? Research on the determinants of savanna vegetation structure can be classified into a series of overlapping dichotomies that emphasize competing forces or processes as key determinants of tree-grass coexistence and tree cover. These include: (1) top-down (mainly fire and herbivory) versus bottom-up (soil moisture and nutrients) factors as dominant ecological forces; (2) competition (niche-based) versus demographic-bottleneck models; (3) spatially-explicit versus spatially-unstructured processes; and (4) deterministic versus stochastic dynamics. Many of these are related: deterministic, niche-based theories can explain tree-grass coexistence through spatially-aggregated mathematical models that give rise to stable equilibria. The competing extremes of these dichotomies may also potentially shift in importance along environmental gradients, the most important of which is rainfall, which divides savannas into "dry" and "wet" savannas. For example, stochasticity in a bottom-up driver (soil moisture) may enhance tree survival in dry savannas, whereas a top-down factor (fire) may suppress trees in wet savannas. Research on these processes across a wide range of savanna ecosystems has advanced our understanding of vegetation determinants in individual cases, but no consensus has yet emerged that allows us understand how individual studies can contribute to explain savanna vegetation patterns on a global scale. This symposium will provide a forum for a much-needed synthesis on savanna theory. The speakers include both empiricists and theorists working on savanna environments spanning a wide range of savanna types across the globe, and whose work emphasizes a wide array of mechanisms and approaches to the savanna question. This will be a highly integrative session designed to shed new light on an old problem through novel empirical and theoretical research.