Savanna communities are characterized by a discontinuous overstory of trees in a flammable graminoid-dominated groundcover or ‘grassland’ matrix. Establishing the theoretical framework for the co-dominance of grasses and trees in savanna communities has been an area of intense research interest. A wide range of theoretical models of savannas have been put forward, ranging from equilibrium models based on niche partitioning to non-equilibrium models that result from repeated disturbances. Niche partitioning models have traditionally predicted coexistence of graminoids and trees based on differential access to soil water. Grasses are hypothesized to be better competitors for shallow soil water, while woody vegetation is proposed to have exclusive access to deep soil water. Disturbance-mediated demographic bottleneck models (hereafter, disturbance models), in contrast, are based on externally imposed limits to the ability of overstory trees to suppress the graminoid understory. Disturbances (e.g., fires, hurricanes, droughts) are hypothesized to depress certain life stages of trees in ways that limit tree density, producing an unsaturated canopy above a graminoid understory. Theoretical models of savannas have begun to incorporate aspects of both equilibrium and non-equilibrium models as well as spatial representations of these processes. Niche partitioning models, for example, have been modified to include the effects of herbivory and fire. Disturbance models have incorporated feedbacks between fire and vegetation, for instance, stabilizing savannas communities in an equilibrial savanna state. The wide variety of theoretical representations of savannas results, in part, from the geographical diversity of savannas: Savannas occur on all continents except Antarctica and are found in environments that range from mesic to arid. In our symposium, we propose to explore a variety of savanna systems in order to identify commonalities and suggest directions for development and synthesis of existing savanna theory. We will thus include speakers whose research reflects the diversity of savanna systems and we will also include time for discussion and synthesis. We have invited speakers with expertise in savannas of the southeastern United States, the southwestern United States, South America, and Africa. Our speakers represent those both with primary interests in theory as well as those more empirically oriented. We are proposing eight speakers each with 20 minutes for their presentation, leaving approximately 35 minutes at the end of our symposium for discussion and synthesis of presentations. We have confirmed the participation of all eight proposed speakers.