Between 1985 and 1987, ten areas of upland, marginal grassland were surveyed for breeding waders; eight areas west of the Pennines in the Eden Valley, Cumbria, and two East of the Pennines at Alston, Cumbria and Teesdale, Co. Durham. Grassland improvement resulted in the virtual disappearance of snipe and marked decreases in both the density and the proportion of fields used by breeding lapwing, curlew and redshank. The absence of snipe following improvement was due to land drainage, whereas reduced curlew densities resulted from vegetation changes, redshank were relatively unaffected provided wet areas persisted nearby. Data on lapwing breeding success were obtained from a study of 637 clutches. Fewer clutches suffered predation on unimproved than on improved pastures and resulted in higher hatching success. This, combined with more failed clutches being replaced and better survival by young chicks, resulted in higher productivity on unimproved than on improved pastures. The role of predation in reducing hatching success was confirmed experimentally. In addition, more clutches were destroyed by agricultural activities on improved than on unimproved meadows, resulting in lower productivity. Productivity on unimproved areas was enough to sustain the population, but was insufficient on improved areas and it is suggested that this, together with high philopatry, causes the decline in density of lapwings on improved areas. Invertebrates were sampled by chemical expellents, soil cores and pitfall traps. Grassland improvement resulted in increased earthworms and beetles other than carabids, but decreased spiders and carabic beetles. Fewer species of spider after improvement were largely due to lowered vegetation architecture diversity, whilst fewer species of tipulids resulted from reduced soil moisture and loss of food plants. Changes in species composition were apparent when comparing the seasonal pattern of pitfall catches on unimproved and improved grassland.