For sustainable logging to be achieved in tropical forests, there must be successful gap-phase regeneration to restock the logged-over area. This study examined three aspects of gap-phase regeneration in selectively logged lowland swamp forest of northeast Costa Rica. First, logging gaps were censused immediately after extraction to determine the density of advanced regeneration. Stem density and basal area of residual trees ≥ 10 cm dbh in logging gaps was >85% lower than undisturbed forest, and all trees in gaps had sustained structural damage. The common canopy species Pentaclethra macroloba (Fabaceae) was the most abundant species in gaps whereas the timber tree Carapa nicaraguensis (Meliaceae) was absent from all censused gaps. This suggests that canopy replacement, particularly by Carapa, will depend on trees <10 cm dbh or by seed input into logging gaps. Second, the diversity of the understorey was compared with 6-y old single-tree and multiple-tree logging gaps. Multiple-tree logging gaps were the most diverse, but dominated by two ruderal species; however many shade-tolerant species were present in those gaps. This indicates that controlled selective logging can result in a localized shift in species composition, but that logging gaps should return to pre-logging composition with time under a carefully implemented, controlled harvesting regime. Finally, this study found a significant effect of a fringing Carapa tree on logging gap seedling density. Thus, seed arrival into gaps is a barrier to logging gap regeneration, particularly for a large-seeded tree species. Gap-phase regeneration by a large-seeded tree species in managed forest would benefit from seed broadcasting into gaps.