Sapling mortality exceeds recruitment for many species of trees in the dry forest of 16-y-old islets in Lago Guri, Venezuela. Failure of sapling recruitment is potentially attributable to the aberrant animal communities of these islands. Predators of vertebrates are absent and densities of pollinators and seed dispersers are substantially reduced in comparison to the nearby mainland. In contrast, predators of invertebrates, rodents and generalist herbivores (leaf-cutter ants, howler monkeys, common iguanas) are present at greatly elevated densities. Given these distortions in the animal community, recruitment failure of saplings could potentially be attributable to several causes: reduced pollination or seed dispersal, excessive seed predation or seedling herbivory. Two of these hypotheses are tested herein. The seed predation hypothesis predicts greater seed removal in the presence of hyperabundant rodents, and the seedling herbivory hypothesis predicts reduced seedling survival in the presence of hyperabundant herbivores. Seed removal trials were conducted with 18 species of forest trees, using both exposed and lightly buried seeds. Seed removal was not generally higher on islands supporting hyperabundant rodents, contrary to the seed predation hypothesis. Seedlings exposed to herbivores for 4 mo suffered the highest mortality on small islands supporting hyperabundant rodents and leaf-cutter ants. Saplings survived equally well in cages open to arthropods + rodents as in cages open only to arthropods, suggesting that there was no additive effect of rodents on seedling mortality. Current evidence points to excessive seedling herbivory by arthropods, especially by leaf-cutter ants, as the principal cause of recruitment failure on predator-free Lago Guri islets.