In the northern Great Plains, native prairies managed by the U.S. Fish and Wildlife Service (Service) can be pivotal in conservation of North America's biological diversity. From 2002 to 2006, we surveyed 7,338 belt transects to assess the general composition of mixed-grass and tallgrass prairie vegetation across five "complexes" (i.e., administrative groupings) of national wildlife refuges managed by the Service in North Dakota and South Dakota. Native grasses and forbs were common (mean frequency of occurrence 47%-54%) on two complexes but uncommon (4%-13%) on two others. Conversely, an introduced species of grass, smooth brome (Bromus inermis), accounted for 45% to 49% of vegetation on two complexes and another species, Kentucky bluegrass (Poa pratensis) accounted for 27% to 36% of the vegetation on three of the complexes. Our data confirm prior suspicions of widespread invasion by introduced species of plants on Service-owned tracts of native prairie, changes that likely stem in part from a common management history of little or no disturbance (e.g., defoliation by grazing or fire). However, variability in the degree and type of invasion among prairie tracts suggests that knowledge of underlying causes (e.g., edaphic or climatic factors, management histories) could help managers more effectively restore prairies. We describe an adaptive management approach to acquire such knowledge while progressing with restoration. More specifically, we propose to use data from inventories of plant communities on Service-owned prairies to design and implement, as experiments, optimal restoration strategies. We will then monitor these experiments and use the results to refine future strategies. This comprehensive, process-oriented approach should yield reliable and robust recommendations for restoration and maintenance of native prairies in the northern Great Plains. 2009 by the Board of Regents of the University of Wisconsin System.