Selective logging has degraded large areas of tropical forest. However, disturbed forests are important refugia for biodiversity and recover biomass over time, especially when aided by active restoration. Yet we know relatively little about the effects of logging and restoration on long-term seedlings dynamics and future tree communities. Here, I investigate differences in forest structure and tree community composition at different life-stages across unlogged, naturally regenerating (NR) and actively restored (AR) forests in Danum Valley, Malaysian Borneo, 23-39 years after logging and 12-28 years after restoration. Basal area (all stems ≥2 cm DBH) was greatest in unlogged and did not differ between NR and AR forests. However, mean DBH was greater in AR than NR forest, suggesting that restoration accelerated the return of larger stems. Species diversity was equal across forest types but community composition differed and did not converge over time since logging. After mast fruiting in 2019, seedling density was greater in unlogged and AR forests than in NR forest. However, seedling density in AR forest also declined fastest so, after 1.5 years, AR forest supported fewest seedlings – 0.85% of the original cohort. Larger seedlings established before mast (recorded over 2.5 years) also had greatest mortality in AR forest, but had higher relative growth rates. Masted seedling community compositions diverged over time and dipterocarps exhibited particularly high mortality in AR forest. Differences in forest structure, community composition and recruitment processes therefore persist up to 40 years after logging. These findings may result from: differences in seed and seedling predation, viability, and differing microclimates between forests with different disturbance histories. I highlight the need for further observations of early tree life-stages in recovering forests and their drivers, which ultimately constrain the species pool of future canopy communities, and should guide decision-making in forest restoration.