In many biogeochemical models, plant species are aggregated such that only one generic plant type is represented. If multiple species are present, each “species” or functional type is represented as a collection of physiological traits. We recently have been exploring the physiological responses of co-occurring plant species of a variety of growth forms to determine how many separate physiological types are needed in order to capture the dynamics of net primary production, decomposition, carbon (C) storage and nitrogen (N) availability. We measured species-level photosynthetic responses, canopy light extinction, leaf N and lignin, and soil nutrient processes. In both the prairie and the shrub-savanna we found that photosynthetic responses scale with light within the plant canopy such that for calculation of C gain, the canopy can be modeled as a single unit, regardless of the vertical distribution of species. Nutrient cycling and C storage are, however, very different in herbaceous areas compared to wooded areas, such that shrub-dominated areas have different soil C levels and N mineralization rates from grassland areas. In Texas, shrub-dominated areas have higher soil C and N mineralization. The dominant woody plant in the Texas site is an N fixer. Significant areal expansion of this plant, documented since the mid-1800s, has undoubtedly affected regional patterns of N cycling and net primary production (NPP) (similar to results of Vitousek and Walker, 1989, from Hawaii). The dominant woody plant in Kansas is not a fixer and invasion (which is locally common due to fire suppression) results in losses of N availability and stored C. We suggest that some ecosystem processes may be modeled successfully with highly aggregated models, particularly models of processes that are highly constrained by environmental factors. Other processes will require detailed specification of physiological traits at the species or population level. The degree of species aggregation in ecosystem models should be regarded as a research problem rather than a quasi-ideological problem.