Isolated wetlands are often degraded by agriculture, increasing sediment accretion and altering plant composition. Two common opposing wetland restoration practices are self-design vs. intensive revegetation. Self-design restores hydrogeomorphology without inoculating wetland taxa into restoration sites. Self-design may not meet restoration targets if dispersal-limited plants do not colonize restoration sites. Alternatively, intensive revegetation (hydrogeomorphic restoration combined with revegetation) is costly and time consuming. We investigated plant dispersal-limitation in 309 isolated wetlands among two agricultural landscapes in the U.S. Great Plains (the western High Plains (WHP) and the Rainwater Basin (RWB)) and three land-uses (reference, croplands, and previous croplands) to predict optimal restoration practices. We present analytical tools predicting whether self-design or intensive revegetation will be more successful elsewhere. In the WHP and RWB, perennial wetland cover was 61% and 31% greater in reference than in other land-uses. Distance to the nearest reference wetland explained perennial wetland richness in both regions, and area of reference wetlands within 15. km also was important in the WHP. Annual wetland species were over-represented in previous cropland wetlands and were less influenced by landscape isolation. We analytically identified dispersal-limited and cosmopolitan species in reference wetlands, with distance to reference wetlands and area of surrounding reference wetlands important in determining composition. Further, dispersal-limited plants in reference wetlands had greater cover in clustered than isolated wetlands in previous croplands. Plant community patterns in reference conditions may predict composition in restored wetlands. This aids selection of self-design or revegetation restoration approaches for individual plant species in isolated wetlands. (c) 2012 Elsevier Ltd.