Long-term changes in Denny Wood, an ancient wood-pasture in the New Forest, were recorded on two permanent transects. One was located in an unenclosed part of the wood, which remained open to all large herbivores (ponies,deer, cattle). The other was in an enclosed part, which was fenced against livestock (but not deer) from 1870 to ∼1960, after which ponies regained access. Records of trees, shrubs, seedlings and vascular ground vegetation within the transects were made mainly during 1959-1964 and 1996-2001. The unenclosed transect changed greatly. In 1959 it comprised closed beech-oak Fagus sylvatica-Quercus robur forest with abundant holly Ilex aquifolium understorey and small patches of Agrostis lawn or Pteridium aquilinum grazed by moderate numbers of large herbivores. Forty years later this had become an open oak-beech parkland with little understorey. Species-rich lawns and stands of Pteridium had spread extensively, and large herbivores had become far more numerous. Oak survived much better than beech: the larger beech suffered particularly from drought in 1976 and the medium beech from debarking by grey squirrels Sciurus carolinensis. Many beech were cut and removed in a safety/salvage operation in 1981. Holly was reduced mainly by browsing and debarking by ponies and deer. Tree and shrub seedlings became more abundant, but browsing effectively prevented regeneration. The enclosed transect also comprised closed beech-oak-holly forest in 1959. However, the stand was denser and the ground vegetation was more heavily shaded, species-poor and sparser than on the unenclosed transect. Substantial gaps were created during the next 40 years due mainly to the loss of large beech and oak after drought in 1976 and later storms. Rank, species-poor grassland and stands of Pteridium developed mainly below gaps, but many parts remained as closed high forest with little ground vegetation. No dead trees were removed and dead wood volumes remained consistently high. Holly was much reduced, principally due to browsing and debarking by deer and ponies. Grazing almost completely prevented tree regeneration. The implications for grazing of wood-pastures are discussed. Prolonged exclusion of livestock with reduced deer numbers allowed abundant regeneration of trees with holly, but the ensuing dense, closed stand had an impoverished ground flora. Moderate numbers of livestock (both ponies and cattle) and deer kept some grassy glades, but allowed holly to spread. High numbers of livestock (mainly ponies) with deer reduced holly and, where the canopy opened, helped restore herb-rich ground vegetation, aided by the continuity of grazed turf and refuges provided by Pteridium for grazing-sensitive plants. Nevertheless, heavy grazing mainly by ponies prevented tree regeneration for many decades, even in large openings and despite the presence of possible protective features, i.e. spiny bushes, fallen trees, and stands of Pteridium. The widespread failure of regeneration, reduction in underwood and scrub, accelerating mortality and break up of old-growth stands, and damage to young trees by grey squirrels, represents a vulnerable point in the history of New Forest wood-pastures.