Three different approaches for assessing the impact of a permanent increase in water level on wetland vegetation were studied using a long—term, controlled, and replicated experiment. These three approaches were: (1) digitized vegetation maps derived from aerial photographs; (2) vegetation data (species abundance, species diversity) from 10 permanent quadrats in each cell; and (3) Bray—Curtis similarity indices comparing the composition of the vegetation in permanent quadrats within a cell and among cells. This study was conducted in a 10—celled wetland complex in the Delta Marsh, Manitoba, Canada. There were three water level treatments: (1) the normal or mean regulated water level in the surrounding Delta Marsh, (2) the medium (30 cm above normal), and (3) the high (60 cm above normal). There were four, three, and three cells (ranging in total area from 6 to 8 ha), respectively, in each treatment. The vegetation in all cells had been reestablished with a drawdown just prior to this study. To reduce cell—to—cell variation, we adjusted the indicators derived from vegetation maps from 1985 through 1989 by subtracting the mean values of the same indicators in 1979 and 1980, after 15—16 yr of normal water level conditions. The adjusted percentage of a cell covered with open water increased significantly and two other adjusted indicators, the number of vegetation types and the number of multispecies vegetation types, decreased significantly in the flooded treatments. The percentage of a cell covered with sparse emergent vegetation and percentage covered with standing litter did not differ significantly among treatments. In permanent quadrats, species richness, total shoot density of the emergent species, and the Shannon diversity index showed significant treatment effects: all three declined in the flooded treatments. The Simpson's index, however, did not show a treatment effect. When Bray—Curtis similarity indices comparing the vegetation either among permanent quadrats within a cell or for the same permanent quadrat in a cell among years were used, either within—cell vegetation heterogeneity or ongoing successional changes in the vegetation made it impossible to detect treatment effects.