The relative importance of different bottom-up-mediated effects in shaping insect communities in tropical secondary forests are poorly understood. Here, we explore the roles of vegetation structure, forest age, local topography (valley vs. hill top) and soil variables in predicting fruit-feeding butterfly and tree community composition, and tree community composition in predicting fruit-feeding butterfly community composition, in different-aged naturally regenerating and primary forests of Kibale National Park, Uganda. We also examine which variables are best predictors of fruit-feeding butterfly species richness or diversity. Butterflies (88 species) were sampled with a banana-baited trap and trees (98 taxa) with a 40 × 20-m sampling plot at 80 sampling sites. The environmental variables explained 31% of the variation in the tree community composition, the best predictors being local topography, forest age and cover of Acanthus pubescens (a shrub possibly arresting succession). The fruit-feeding butterfly community composition was better predicted by tree community composition (explaining 10% of the variation) rather than vegetation structure, local topography or soil factors. Environmental variables and tree species richness (or diversity) were poor predictors of butterfly species richness (or diversity). Our results emphasize the importance of tree community to recovery of herbivorous insect communities in tropical secondary forests.