Developing restoration strategies that accelerate natural successional processes and are resource‐efficient is critical to facilitating tropical forest recovery across millions of hectares of deforested lands in the tropics.

We compared tree recruitment after a decade in three restoration treatments (natural regeneration, applied nucleation/island tree planting and plantation) and nearby reference forest in the premontane rain forest zone in southern Costa Rica. The study was replicated at 13 sites with a range of surrounding forest cover, enabling us to evaluate the relative influence of local restoration treatments and landscape forest cover on tree recruitment.

Density of small‐seeded (<5 mm), animal‐dispersed recruits was lower in natural regeneration than in applied nucleation, plantation or reference forest plots. Species richness, species density and density of medium (5–10 mm)‐ and large (>10 mm)‐seeded, animal‐dispersed recruits were greatest in reference forest, intermediate in applied nucleation and plantation and lowest in natural regeneration plots.

Recruit composition differed substantially between reference forest and all restoration treatments. In general, plantation recruit composition was more similar to reference forests and natural regeneration least similar; however, there was high within‐treatment variation.

Models suggested weak support for the effect of surrounding forest cover on tropical tree recruit density and composition, as compared to restoration treatment and site conditions (e.g. elevation), in this intermediate forest cover landscape.

Synthesis and applications. Applied nucleation appears to be a cost‐effective strategy as compared to plantation‐style planting to accelerate tropical forest recovery regardless of the amount of forest cover immediately adjacent to the site. However, even with active restoration interventions, forest recovery is a multidecade process that proceeds at highly variable rates.

In June-July 2015 (9-11 years after plot establishment), we sampled tree seedlings (>0.2 and <1 m tall) and saplings (>1 m tall and <5 cm d.b.h.) in all restoration and in reference forests at six sites remnant forest nearby that had not been logged or used for agriculture. We identified saplings in 40, 2 × 4 m quadrats in each plot (320 m2 per plot) and recorded seedlings in 1 × 2 m quadrats nested within the sapling quadrats (80 m2 total). Quadrats were distributed in four belt transects of 10 each in natural regeneration, plantation, and reference forest treatments with one randomly located in each of the four quadrants of the plot. In the applied nucleation treatment, quadrats where distributed along six belt transects aligned from the interior to outside the initial planted areas of the three nuclei sizes, with six adjacent quadrats in small and medium nuclei and seven in large nuclei. The 5-m edge at the perimeter of the plot was not sampled.