Two watersheds on the southern end of Havens State Wildlife Management Area, Roanoke County, Virginia, were sampled for vascular plant species richness. Two-hundred and forty-eight species were identified. Three methods of sampling for species richness in eastern forests were compared: timed-search meanders, belt transects, and plots. Meanders and transects located more species in the same amount of time as plots. Plot sampling encompassed only two-thirds of the richness known from the site. Species-area and species-effort relationships were described by exponential models (number of species = In (area + 1), and number of species = In (time + 1)). Models were used to predict numbers of species which might have been found had more area been sampled or had more time been spent searching. Species-area models yielded more conservative, and probably more accurate, predictions than species-time models. Predictions of species numbers were reasonable for areas as large as 60 ha, but were rather large for areas the size of Havens (2800 ha). Sufficiency of search effort was judged using species-area and species-effort curves. However, it was difficult to tell whether the curves approached horizontal or not. Therefore, this was not a good technique to judge sampling adequacy. Species composition observed by each of the three methods was different. Composition of species lists was 65% similar between meanders and transects, and only 51-58% similar between plots and other methods. These figures were within an expected range. A combination of two methods or repeated meanders was recommended in order to identify a higher proportion of the species present. Seasonal and observer differences, and the effect of learning and taxonomy on richness estimates were discussed. / Master of Science