Understanding herbivore optimization has implications for theories underpinning ecosystem processes, management of large herbivores, and the landscapes they inhabit. We designed an experiment to examine interactions related to density dependence of North American elk (Cervus elaphus) and resulting plant responses to herbivory in the Blue Mountains of Oregon, USA, from 1999 to 2001. We experimentally created high (20.1 elk/km²) and low (4.1 elk/km²) population densities of elk and built exclosures to examine effects of herbivory on productivity and species composition of plants. We hypothesized that if herbivore optimization occurred with increasing density of elk, there should be a concordant increase in plant production, followed by a decline in productivity as grazing intensity continued to increase (i.e., herbivore optimization). Net aboveground primary productivity (NAPP) increased from no herbivory to herbivory by elk at moderate density and then declined as herbivory by elk continued to increase in areas with high NAPP (mesic and logged forests) but not in areas with low NAPP (xeric forests and grasslands). Herbivore optimization occurred across all functional groups of plants, including graminoids, forbs, and shrubs for high-NAPP areas. Herbivore optimization may be difficult to detect in woody plants compared with graminoids because of their differing structure and growth forms. Although herbivore optimization previously has been reported in grasslands, our study documents this phenomenon in woody plant communities. We hypothesize that such subtle changes in NAPP from herbivory might be more common than previously thought; carefully designed experiments are required to detect those responses to herbivory by large herbivores. Apparent offtake of plants followed a similar pattern to NAPP and was greatest at intermediate levels of herbivory by elk, and then declined as NAPP approached zero. Quality of plants, as indexed by percent nitrogen (N), also exhibited a parabolic function with increasing density of elk. Nonetheless, we observed no changes in species composition or diversity of plants with our density manipulations of elk, probably because of the extensive history of grazing by native and domestic herbivores in the Blue Mountains, the resilience of the remaining plants to herbivory, and the short 3-year duration of our study. Likewise, we observed no increases in rates of nutrient cycling with changes in densities of elk, perhaps because areas where large amounts of elk urine and feces were concentrated (e.g., grazing lawns) did not occur in this ecosystem.

Herbivore optimization could have ramifications for population dynamics of ungulates. We argue that other processes, such as migration, predation, or harvest, might be necessary to maintain areas of increased NAPP because of strong density-dependent feedbacks and negative effects of ungulates on their food supply would not allow equilibrium between herbivores and their food supply at the low densities where we documented herbivore optimization. We also hypothesize that increased NAPP at low to moderate population densities of herbivores could lead to rapid population growth of these herbivores and help reduce density-dependent feedbacks and promote overshoots of carrying capacity. Our data do not support stocking large densities of ungulates on western rangelands to obtain peak production of NAPP. We recommend maintaining low to moderate densities of large herbivores in ecosystems if goals are to maximize NAPP and forage quality or to maximize body condition and reproduction of ungulates.