Past studies have shown that tropical forest regeneration on degraded farmlands is initially limited by lack of seed dispersal, but few studies have tracked changes in abundance and composition of seed rain past the first few years after land abandonment. We measured seed rain for 12 months in 10 6–9‐year‐old restoration sites and five mature, reference forests in southern Costa Rica in order to learn (1) if seed rain limitation persists past the first few years of regeneration; (2) how restoration treatments influence seed community structure and composition; and (3) whether seed rain limitation is contingent on landscape context. Each restoration site contained three 0.25‐ha treatment plots: (1) a naturally regenerating control, (2) tree islands, and (3) a mixed‐species tree plantation. Sites spanned a deforestation gradient with 9–89% forest area within 500 m around the treatment plots. Contrary to previous studies, we found that tree seeds were abundant and ubiquitous across all treatment plots (585.1 ± 142.0 seeds·m⁻²·yr⁻¹ [mean ± SE]), indicating that lack of seed rain ceased to limit forest regeneration within the first decade of recovery. Pioneer trees and shrubs comprised the vast majority of seeds, but compositional differences between restoration sites and reference forests were driven by rarer, large‐seeded species. Large, animal‐dispersed tree seeds were more abundant in tree islands (4.6 ± 2.9 seeds·m⁻²·yr⁻¹) and plantations (5.8 ± 3.0 seeds·m⁻²·yr⁻¹) than control plots (0.2 ± 0.1 seeds·m⁻²·yr⁻¹), contributing to greater tree species richness in actively restored plots. Planted tree species accounted for <1% of seeds. We found little evidence for landscape forest cover effects on seed rain, consistent with previous studies. We conclude that seed rain limitation shifted from an initial, complete lack of tree seeds to a specific limitation on large‐seeded, mature forest species over the first decade. Although total seed abundance was equal among restoration treatments, tree plantations and tree islands continued to diversify seed rain communities compared to naturally regenerating controls. Compositional differences between regenerating plots and mature forests suggest that large‐seeded tree species are appropriate candidates for enrichment planting.

Seed rain was collected twice per month from 15 Feb 2012 ñ 14 Feb 2013 in 420 0.25-m2 traps. In each restoration and reference plot, we established four trapping stations along permanent vegetation transects (Fig. B1). Each station had three traps. In island plots, we placed two of the twelve traps (16%) in the planted patches and the remainder in unplanted areas to match the proportion of the area that was initially planted (18%). All traps were located >5 m from the plot edge and restoration plots were separated by =5 m.