Editor: Michele R. DudashPremise of research. Pollen-dispersal distances may be flexible and dependent on a variety of landscape factors. The study of the effect of plant density and forest fragmentation on dispersal distances is important to understand the response of plant populations to habitat loss or harvesting, especially in tropical timber tree species. Methodology.A comparison was made of the pollen-dispersal curves of the timber tree Carapa nicaraguensis in three different habitats in southern Costa Rica: (i) forest remnants in hills with low density of the study species, (ii) a gallery forest in a swampy area with high density of the study species, and (iii) isolated trees in pastures. Microsatellite genotypes were used to determine the genealogical relationships between adults, seeds, and seedlings collected below tree crowns. Bayesian methods were used to simultaneously estimate the dispersal-distribution parameters, genealogical relationships, genotyping error rates, and numbers of unsampled males.Pivotal results. The pollen-dispersal distances were influenced primarily by the surrounding matrix and degree of spatial isolation of the local population. Pollen dispersal was more extensive between trees in forest fragments than in other habitats. Predominant selfing was found in a single tree isolated in pastures. Evidence of seed dispersal to a different tree crown was found for trees in forest fragments.Conclusions. The results are consistent with those of previous studies indicating that pollen-dispersal patterns are dependent on the degree of isolation and tree density of local populations. Genetic connectivity across the fragmented landscape is maintained by trees in low-density forest fragments and isolated trees in pastures, whereas trees in high-density gallery forests are genetically isolated from the rest of the population.