Understanding tropical forest tree diversity has been a major challenge to ecologists. In the absence of compensatory mechanisms, two powerful forces, drift and competition, are expected to erode diversity quickly, especially in communities containing scores or hundreds of rare species. Here, I review evidence bearing on four compensatory mechanisms that have been subsumed under the terms "density dependence" or "negative density dependence": (1) intra- and (2) interspecific competition and the action of (3) density-responsive and (4) distance-responsive biotic agents, as postulated by Janzen and Connell. To achieve ontological integration, I examine evidence based on studies employing seeds, seedlings, and saplings. Available evidence points overwhelmingly to the action of both host-generalist and host-restricted biotic agents as causing most seed and seedling mortality, implying that species diversity is maintained via top-down forcing. The overall effect of most host-generalist seed predators and herbivores is to even out the distribution of surviving propagules. Spatially restricted recruitment appears to result mainly, if not exclusively, from the actions of host-restricted agents, principally microarthropods and fungi, that attack hosts in a distance-dependent fashion as Janzen and Connell proposed. Near total failure of propagules close to reproductive conspecifics ensures that successful reproduction occurs through a scant rain of dispersed seeds. Densities of dispersed seeds and seedlings arising from them are so low as to generally preclude the operation of density dependence, at least during early ontogenetic stages. I conclude that Janzen and Connell were essentially correct and that diversity maintenance results from top-down forcing acting in a spatially nonuniform fashion.