Dryland ecosystems cover over 41% of the earth’s land surface, and living within these important ecosystems are approximately 2 billion people, a large proportion of whom are subsistence agropastoralists. Improper grazing in drylands can negatively impact ecosystem productivity, soil conservation, hydrologic processes, downstream water quantity and quality, and ultimately human health and economic well-being. Concerns regarding the degraded state of western US rangelands in the 1950s resulted in an interagency committee to study the effects of land use on runoff and erosion processes. In 1953, a federal research group established four paired watersheds in western Colorado to study the interaction of grazing by domestic livestock, runoff, and sediment yield. Exclusion of livestock from half of the watersheds dramatically reduced runoff and sediment yield after the first 10 yr—primarily due to changes in ground cover but not vegetation. Here, we report results of repeated soils and vegetation assessments of the experimental watersheds after more than 50 yr of grazing exclusion. Results show that many of the differences in soil conditions between grazed and ungrazed watersheds observed in the 1950s and 1960s were still present in 2004, despite reduced numbers of livestock: few differences in vegetation cover but large differences in biological soil crusts, soil stability, soil compaction, and soil biogeochemistry. There were differences among soil types in response to grazing history, especially soil lichen cover and soil organic matter, nitrogen, and sodium. Comparisons of ground cover measured in 2004 with those measured in 1953, 1966, and 1972 suggest much of the differences between grazed and ungrazed watersheds likely were driven by high sheep numbers during droughts in the 1950s. Persistence of these differences, despite large reductions in stocking rates, suggest the combination of overgrazing and drought may have pushed these salt desert ecosystems into a persistent, degraded ecological state.