The dominant models explaining tree species diversity and distribution patterns in tropical forests are the Janzen-Connell and Recruitment Limitation models, neither of which considers the effect of long-distance seed dispersal on seed survival, seedling establishment, or the aggregated distributions of trees empirically observed at mesoscales in tropical forests. At a neotropical forest site (Maracá Island Ecological Reserve, Roraima, Brazil), we experimentally reproduced long-distance clumped seed dispersal by tapirs for the palm Maximiliana maripa. Such dispersal protects seeds from attack by species-specific bruchid beetles by (1) covering them in protective fecal material and (2) placing them in sites distant from conspecific adult tree aggregations, where beetles are less active. Endocarps distant from parent patches survived bruchid attack at a significantly higher rate than those in parent patches, as did in-feces endocarps compared to clean endocarps. A significant interaction effect between distance from patches and feces treatment showed that feces conferred protection to seeds within a parent patch but did not appear to confer additional protection to seeds already protected by distance from the parent patch. A mesoscale map compiled from aerial photography, satellite imagery, and air- and ground-truthing revealed an aggregated pattern of M. maripa palms associated with tapir latrine sites, supporting the view that long-distance seed dispersal by tapirs is responsible for the generation of palm patches and potentially important in forest–savanna boundary dynamics. We conclude that seed shadows and survival rates can justifiably be studied at the scale of tree aggregations rather than at the scale of individual trees, and that long-distance seed dispersal is neither rare nor unpredictable once we understand the movements and behavior of large, mobile animals.