1. Marginal agricultural land, which in the UK refers to upland grazings in particular, is going to see changes in management driven by markets, subsidies, grants and environmental change with implications for biodiversity. Using a large-scale, long-term grazing experiment in the UK uplands we assessed the impact of intensification (tripling sheep numbers), abandonment (removal of sheep) and grazer diversification (partial replacement of sheep by cattle) on vegetation composition in a heterogenous area of grassland. 3. Species benefiting from increased grazing included sweet vernal grass (Anthoxanthum odoratum), mat grass (Nardus stricta) and deer grass (Trichophorum cespitosum). Species that benefitted from the removal of grazing included bog asphodel (Narthecium ossifragum), bracken (Pteridium aquilinum) and blaeberry (Vaccinium myrtillus). 4. Responses differed between vegetation communities; more productive acid grassland communities showed little change when grazing was removed, whilst less productive mire communities contained species, capable of increasing after grazing removal. Increased grazing and, to a lesser extent, the introduction of cattle increased species diversity. 5. Synthesis and applications. Marginal agricultural grassland will likely see management change driven by markets, subsidies and environmental change. Vegetation change in these relatively infertile grasslands is slow and features shuffling dominance amongst species in the initial vegetation. Initial structural changes affect other trophic levels in this experiment; the slow change in composition will affect the system over longer timescales. Management decisions in the uplands encompass complex trade-offs between production, biodiversity and a range of ecosystem services. Predicting the consequences of decisions is difficult given the slow dynamics of unproductive habitats.