Wind disturbance profoundly shapes temperate forests but few studies have evaluated patterns and mechanisms of long-term forest dynamics following major windthrows. In 1990, we initiated a large hurricane simulation experiment in a 0.8-ha manipulation (pulldown) and 0.6-ha control area of a maturing Quercus rubra--Acer rubrum forest in New England. We toppled 276 trees in the pulldown, using a winch and cable, in the northwesterly direction of natural treefall from major hurricanes. Eighty percent of canopy trees and two-thirds of all trees > or = 5 cm dbh (diameter at breast height) suffered direct and indirect damage. We used 20 years of measurements to evaluate the trajectory and mechanisms of forest response after intense disturbance. Based on the patch size and disturbance magnitude, we expected pioneer tree and understory species to drive succession. The first decade of analyses emphasized tree seedling establishment and sprouting by damaged trees as the dominant mechanisms of forest recovery in this extensive damaged area. However, despite 80% canopy damage and 8000-m2 patch size, surviving overstory and advance regeneration controlled longer-term forest development. Residual oaks make up 42% of stand basal area after 20 years. The new cohort of trees, dominated by black birch advance regeneration, contributes 30% of stand basal area. There were shifts in understory vegetation composition and cover, but few species were gained or lost after 20 years. Stand productivity rebounded quickly (litterfall recovered to pre-disturbance levels in six years), but we predict that basal area in the pulldown will lag behind the control (which gained 6 m2/ha over 20 years) for decades to come. This controlled experiment showed that although the scale and intensity of damage were great, abundant advance regeneration, understory vegetation, and damaged trees remained, allowing the forest to resist changes in ecosystem processes and invasion by new species.