Abstract: Inventory and classification of natural vegetation is necessary for adequate preservation of representative examples. Savannas of the southeastern coastal plain have not been carefully examined over a state-wide area or larger to determine floristic and environmental relationships. To meet this need, savannas were inventoried in the coastal plain of the Carolinas north of the Congaree-Cooper River System, South Carolina. Savannas were selected that contained a naturally established pine canopy of 50% or less cover, shrub cover of 10% or less, and graminoid-dominated herb cover of 50% or greater. No savannas were observed north of the Neuse River, North Carolina. Within the remaining portion of the study area, forty-eight plots were established to sample tree, shrub, and herb strata. Environmental factors examined included twelve soil parameters, distance from the coast, fire frequency, and topography. DCCA and DCA ordinations and TWINSPAN classification were used to characterize the variation in savanna herb composition. Convergence of plot groupings among the results was used to classify savannas into five communities: wet Ultisol, mesic Ultisol, dry Ultisol, wet Spodosol, and dry Spodosol. Soil moisture and texture were the factors most highly correlated with composition. DCA was used to examine the influence of scale on vegetation pattern, variation within communities and plots, and compositional relationships of savannas throughout the southeastern states. No significant difference in plot composition of communities was observed when herb data representing four-nineteenths of the original sample area were ordinated. Differences in graminoid cover accounted for most of the variation within wet to mesic communities, while drier communities varied more in species richness. Savannas of the Carolinas were similar in composition to samples from Virginia, South Carolina, Georgia, and Florida. Priorities for savanna acquisition may be determined by examination of ordinated herb samples from extant preserves and other savannas in the study area. Field observations and literature indicate that savanna composition and stability depend on fire regime and that preservation requires prescribed burning.