AbstractEcological restoration of abandoned, formerly forested farmland can improve the delivery of ecosystem services and benefit biodiversity conservation. Restoration programs can involve removing isolated, non‐native trees planted by farmers for fruit or wood. As such “legacy” trees can attract seed dispersers and create microclimates that help native seedlings to establish, removing them may actually slow forest recovery. Working on abandoned farmland in Kibale National Park, Uganda, we evaluated the effect of legacy trees on forest recovery by measuring the number, diversity, and biomass of native seedlings and saplings regenerating in plots centered on avocado (Persea americana), mango (Mangifera indica), and Eucalyptus legacy trees compared with adjacent plots without legacy trees. The assemblages of native, forest‐dependent tree species in plots around avocado and mango trees were distinct from each other and from those around eucalyptus and all the near‐legacy plots. In particular, avocado plots had higher stem density and species richness of forest‐dependent species than near‐avocado plots, particularly large‐seeded, shade‐tolerant, and animal‐dispersed species—key targets of many restoration plans. Furthermore, many of the species found in high numbers were among those failing to establish in ongoing large‐scale forest restoration in Kibale. Taken together, our results demonstrate that the legacy trees facilitate the dispersal and establishment of native tree species. Retaining the existing legacy trees for a number of years could usefully complement existing management strategies to restore more biodiverse native forest in degraded lands. However, careful monitoring is needed to ensure that the legacy trees do not themselves establish.