Many pioneer tropical trees have high outcrossing rates. Inbreeding occurs primarily by matings between neighboring relatives. Consequently the manner in which inbreeding levels and heterozygosity change through decreases in effective population size, which may arise, for example, from habitat fragmentation, depends primarily on the degree of spatial structure. Spatial structure could also influence populations' adaptive responsiveness to microenvironmental selection (Epperson 1992). Yet it has been underappreciated that genetic structure should also depend on the life history and life stage of individuals. This paper investigates the importance of seed dispersal on the population genetics and structure of Cecropia obtusifolia, a Neotropical pioneer rainforest tree that is outcrossing and wind pollinated, with seeds that are dispersed by various frugivores. The results show that the type of dispersal system of C. obtusifolia causes marked genetic spatial autocorrelations among the often large numbers of seedlings concentrated within canopy gaps. However, this structure persists in a much weaker relict form in juveniles and adults, probably because by then competition has removed most individuals. The structure of genetic variation has a number of implications for ecological and evolutionary genetics and conservation biology. The level of spatial structure both influences and is influenced by most population genetic factors (Wright 1978). In plant populations, it is intimately connected to levels of pollen and seed dispersal; and studies of spatial structure can be used to estimate dispersal levels (e.g., Epperson 1993). Levels of inbreeding, inbreeding depression, and population survival all can depend on the form and magnitudes of spatial structure of genetic variation (e.g., Wright 1978). It has often been predicted that plant populations with limited seed dispersal should exhibit marked spatial structure even if pollen disperses long distances (e.g., Libby et al. 1969). Yet, in many cases structure has not been detected using Fstatistics. More specifically, genetic correlations among progeny, clustered near their maternal parent because of limited seed dispersal, are predicted to result in substantial genetic structure. However, striking changes in structure can occur as the progeny age, because of a number of often uncited life-history considerations. Primary among these is that many of the correlations among progeny are due to the progeny occuring in much higher densities than the parents. The correlations may greatly diminish as the progeny mature and are thinned out (Epperson 1992). For example, in species (like C. obtusifolia) where seedling recruitment occurs primarily from a single or few age classes and only a small proportion of seedlings survive to replace adults, this kind of structure would essentially disappear. In such cases, and where either pollen or seed disperses relatively long distances, there should be little build up of isolation by distance, even in large continuous populations (Wright 1946). Structure has been found among young progeny of remaining trees (with wind-dispersed pollen) following harvest removal of (long-lived) temperate forest tree species (Neale and Adams 1985; Yazdoni et al. 1985; Furnier et al. 1987), and similar progeny clustering was found in progeny groups of tropical tree species (Sakai 1985). In contrast, little evidence of structure is found among mature adults (Epperson and Allard 1989; Knowles 1990; Schnabel and Hamrick 1990; Perry and Knowles 1991). These results have been sometimes viewed as discrepant (e.g., Schnabel and Hamrick 1990; Perry and Knowles 1991). As noted above, such apparent discrepancies can be explained simply by progeny thinning. However, the forms of structure for two or more life stages are rarely analyzed in the same populations or even in the same species (for an exception see Linhart et al. 1981). It should be noted that strong spatial structure is detected in adult individuals in plant populations with low pollen and seed dispersal (e.g., Schaal 1975; Epperson and Clegg 1986; Schoen and Latta 1989). Cecropia obtusifolia is a Neotropic pioneer tree species that is an important component of subtropical forest canopy. It regenerates strictly in recent gaps in the rainforest canopy. One population of C. obtusifolia has been extensively mapped and studied for allozyme variation, including large amounts of mapped genotypic data for a number of life-cycle stages, including adults, juveniles, and seedlings. Population data from multiple life stages of C. obtusifolia (Alvarez-Buylla et al. 1996) afford an exceptional opportunity to measure and examine the degree of progeny clustering resulting from correlated seed dispersal, and details of the dispersal and structure-forming processes. In this paper we also test the limits to structure detectability, by utilizing very sensitive spatial autocorrelation measures known as join-count statistics (e.g., Epperson 1993) for such structures at various life stages.