1. The 20th century has witnessed substantial increases in the intensity of agricultural land management, much of which has been driven by policies to enhance food security and production. The knock-on effects in agriculturally dominated landscapes include habitat degradation and biodiversity loss. We examine long-term patterns of agricultural and habitat change at a regional scale, using the Peak District of northern England as a case study. As stakeholders are central to the implementation of successful land-use policy, we also assess their perceptions of historical changes. 2. In the period 1900 to 2000, there was a fivefold rise in sheep density, along with higher cattle density. We found a reduction in the number of farms, evidence of a shift in land ownership patterns, and increased agricultural specialization, including the virtual disappearance of upland arable production. 3. Despite previous studies showing a substantial loss in heather cover, we found that there had been no overall change in the proportion of land covered by dwarf shrub moor. Nonetheless, turnover rates were high, with only 55% of sampled sites maintaining dwarf shrub moor coverage between 1913 and 2000. 4. Stakeholders identified many of the changes revealed by the historical data, such as increased sheep numbers, fewer farms and greater specialization. However, other land-use changes were not properly described. For instance, although there had been no overall change in the proportion of dwarf shrub moor and the size of the rural labour force had not fallen, stakeholders reported a decline in both. Spatial heterogeneity of the changes, shifting baselines and problems with historical data sources might account for some of these discrepancies. 5. Synthesis and applications. A marked increase in sheep numbers, combined with general agricultural intensification, have been the dominant land-use processes in the Peak District during the 20th century. Stakeholders only correctly perceived some land-use changes. Policy and management objectives should therefore be based primarily on actual historical evidence. However, understanding stakeholder perceptions and how they differ from, or agree with, the available evidence will contribute to the successful uptake of land management policies and partly determine the costs of policy implementation.