Little experimental information is available on the relationship between herbivory by native ungulates and vegetation in relatively undisturbed environments. A quasi-experimental situation exists in Rocky Mountain National Park, where elk (Cervus elaphus) populations have increased about 3-fold since 1968, following their release from artificial controls within the park boundaries. We reviewed data collected on vegetation transects established and monitored over the 25-year period from 1968 through 1992. Data were subjected to rigorous statistical analysis to detect trends following the release of elk from artificial controls. Increases in elk habitat use and decreases in deer habitat use were observed on all transects over the 25-year period. Significant increases in moss and lichen cover occurred in three offour vegetation types. Percent cover of bare ground, forbs (particularly Selaginella densa), and Carex spp. increased on grassland transects. Increases in timothy (Phleum pratense) were observed on meadow transects. Graminoid and litter cover increased on sagebrush transects, and shrub and litter cover increased on bitterbrush transects.

We concluded the lack of control (fenced) plots in this sampling design, the types ofmeasures, the small number of replicates, and nonrandom placement of plots limit the inferences and sensitivity from the work. Unique strengths ofthe work included the long time period (25 years), good distribution of samples, consistency ofthe observer (D. Stevens), and placement of the plots in the most heavily grazed sites.

Some grazing-induced responses were detected. Grazing-resistant species such as sedges (native), timothy (exotic), and club mosses increased and the amount of bare ground increased on some grazed sites. However, the changes within this sampling program alone were not alarming. The amount of bare ground increase was minor (4%), and grass and shrub cover increased in the shrub plots. The inferential power ofthis sample design was limited to the study plots only. Other factors (climate change, succession) were not controlled for using fenced plots and the sensitivity ofthe methods and plots to detect change were limited. For example, the low number oftransects in willow was not adequate to monitor conditions on the entire winter range. Lacking controls, observed changes may have been due to other factors (climate trends, beaver dam abandonment, stream channel changes), not elk herbivory alone. We recommend using a new sampling design that would include controls, pretreatment data, random site selection, and much more replication.