Although free-roaming equids occur on all of the world's continents except Antarctica, very few studies (and none in the Great Basin, USA) have either investigated their grazing effects on vegetation at more than one spatial scale or compared characteristics of areas from which grazing has been removed to those of currently grazed areas. We compared characteristics of vegetation at 19 sites in nine mountain ranges of the western Great Basin; sites were either grazed by feral horses (Equus caballus) or had had horses removed for the last 10-14 years. We selected horse-occupied and horse-removed sites with similar aspect, slope, fire history, grazing pressure by cattle (minimal to none), and dominant vegetation (Artemisia tridentata). During 1997 and 1998, line-intercept transects randomly located within sites revealed that horse-removed sites exhibited 1.1-1.9 times greater shrub cover, 1.2-1.5 times greater total plant cover, 2-12 species greater plant species richness, and 1.9-2.9 times greater cover and 1.1-2.4 times greater frequency of native grasses than did horse-occupied sites. In contrast, sites with horses tended to have more grazing-resistant forbs and exotic plants. Direction and magnitude of landscape-scale results were corroborated by smaller-scale comparisons within horse-occupied sites of horse-trail transects and (randomly located) transects that characterized overall site conditions. Information-theoretic analyses that incorporated various subsets of abiotic variables suggested that presence of horses was generally a strong determinant of those vegetation-related variables that differed significantly between treatments, especially frequency and cover of grasses, but also species richness and shrub cover and frequency. In contrast, abiotic variables such as precipitation, site elevation, and soil erodibility best predicted characteristics such as forb cover, shrub frequency, and continuity of the shrub canopy. We found species richness of plants monotonically decreased across sites as grazing disturbance increased, suggesting that either the bell-shaped diversity-disturbance curve of the intermediate-disturbance hypothesis does not apply in this system or that most sites are already all on the greater-disturbance slope of the curve. In our study, numerous vegetation properties of less-grazed areas and sites differed notably from horse-grazed sites at local and landscape scales during a wetter and an average-precipitation year. ?? 2007 Springer Science+Business Media B.V.