When compared to planted reforestation, natural unassisted regeneration is often reported to result in the slow recovery of biomass and biodiversity, especially early in succession. In some cases, naturally regenerating forests are not comparable to the community structure of primary forests after many decades. However, direct comparison of the outcomes of tropical forest restoration and natural regeneration is hindered by differences in metrics of forest recovery, inconsistency in land-use histories, and dissimilarities in experimental design. We present the results of a replicated reforestation experiment comparing natural regeneration and polyculture tree planting at multiple diversity levels (3, 6, 9, or 12 native tree species), with uniform land-use history and initial edaphic conditions. We compare the recovery of basal area and floristic diversity in these treatments after 5 yr of succession. Total basal area was higher in planted plots than in naturally regenerating plots, but it but did not vary among the different planted diversity levels. The basal area of woody recruits did not differ among treatments. The diversity of woody recruits increased substantially over time but did not vary among planting treatments. Species composition trajectories showed directional turnover over time, with no consistent differences among treatments. The convergence of restoration trajectories and similarity of floristic community diversity and composition across all treatments, after only 5 yr, provides evidence of the viability of natural regeneration for rapid restoration of forest biodiversity.