To examine how selection may act on seed dispersal, we experimentally manipulated seed distributions and compared levels and patterns of seedling recruitment in three altered distributions, relative to the nonmanipulated, leptokurtic distribution. We picked up, then redistributed the wind—dispersed seeds of two isolated parents of Tachigalia versicolor, tropical tree species growing on Barro Colorado Island (BCI), Panama. Three experimental distributions were created in long columns adjacent to or well beyond the undisturbed column of the natural distribution: (1) a nonleptokurtic "even" distribution of uniform density equal to the mean density throughout the natural column, (2) a "mixed" distribution with the same total number and kurtosis as the natural column, but composed of an equal proportion of seeds from the two parent trees, and (3) an "extended—tail" distribution spread in low uniform density up to 1.8 km beyond the natural distribution. Thus, the study addressed the relative importance of density, distance, and genetic makeup on offspring recruitment. The results of the 2 yr observations of seed and seedling survival indicate that, relative to the natural distribution, seedling recruitment at 2 yr was greater in the even distribution, was minimally affected by a mixture of genotypes from overlapping distributions of two parent trees, and was very infrequent from seeds dispersed to a wide variety of distant sites. Although the mixture of genotypes from overlapping distributions did enhance the probability of survival at one site, that effect occurred only during the seedling and not the seed stage. For the extended—tail and mixed distributions, in which offspring of both parents occurred at both sites, the impact of site differences (survival in Site A tail > Site B tail; survival in Site A mixed > Site B mixed) exceeded that of parent differences (Parent A > Parent B). The major differences among distributions arose in the seed and early seedling stages (2 experimental quadrats was limited to very high densities and occurred only in early seedling stages (0—2 mo). Distance effects on the probability of survival were inconsistent both between sites and among distributions. However, the absolute number of seedlings surviving at 2 yr declined with distance from the parent in all three distributions, except the even distribution at Tree A. Additionally, seedling recruitment at any distance 2). Therefore, we conclude that selection on dispersal by Tachigalia versicolor on BCI should act to increase the uniformity of the seed distribution, but not to increase dispersal distance to the point that the density of the seed shadow is diluted.