Predictive modeling of plant distributions rests on the assumption that correlations exist between the presence/absence of a species and selected climate, topographic, substrate, and land cover variables. Once these underlying patterns are determined, maps can be created in GIS that identify all areas that meet the specific conditions for a given species. Such maps can be used to prioritize areas for field surveys of rare plants or assist decision makers in project clearance activities. Using classification tree analysis, we developed correlational models for 44 Wyoming plant species listed as BLM Sensitive or Threatened or Endangered under the Endangered Species Act. Presence/absence of each species was the response variable in the models and was derived from location records of the Wyoming Natural Diversity Database and Rocky Mountain Herbarium. Environmental variables, including total monthly precipitation, average monthly air temperature, monthly shortwave radiation, number of wet days, growing degree-days, local topographic relief, bedrock and surficial geology, soils, elevation, and land cover, were used as predictors. Location data were randomly subdivided into model-building and validation data sets to test the classification success of the final models. Species with fewer than 16 present points were also modeled using the range/intersection method in which the range of environmental values at all present sites of a species were intersected in GIS to identify areas with similar attributes across the state. Wetland plants were modeled with classification tree or range/intersection methods and the resulting models were then overlaid with a riparian/aquatic model to highlight suitable wetland areas within the species' predicted range. We found that the distribution of rare species in Wyoming was most strongly correlated with specific bedrock and soil types, but was also influenced by topographic relief, land cover, and various monthly precipitation and temperature values. Overall, our models were conservative in the area predicted for these species and typically had low false positive or commission error rates. Due to the limited number of samples available, we were unable to determine the false negative or omission error rates with validation data for many of the plant species. For those that could be tested, the omission error rates were moderate to high. The distribution maps produced by correlational modeling did an excellent job of identifying areas where rare species are unlikely to occur and did a good job of highlighting areas of potential habitat that warrant additional on-the- ground survey.