Summary1. The term ‘dispersal limitation’ represents two distinct component processes: the number of seeds produced (fecundity) and the spatial pattern of the seed rain (distribution). We present a quantitative evaluation of these component processes of dispersal limitation for a tropical forest tree community.2. Using a regularly spaced grid of 289 seed traps (0.5 m2 each), we monitored the seed rain into 1.44 ha of upper Amazonian floodplain forest for 6 years whilst concurrently monitoring sapling recruitment in a 0.81-ha subplot centred within the seed-trapping grid. This arrangement allowed us to compare the spatial pattern of seed rain with that of sapling recruitment.3. We endeavoured to distinguish between undispersed and dispersed seeds by applying a series of criteria to seeds collected in the traps and by removing from certain analyses all seeds that fell under reproductive conspecifics. Gross fecundity of 30 common species that contribute to the advanced regeneration was uniformly low and the rain of dispersed seeds was lower still, being <1.0 m−2 year−1 in every case.4. The rain of dispersed seeds with respect to conspecific reproductives closely matched the recruitment of saplings, whereas gross seed rain (all seeds, including undispersed seeds) did not.5. Synthesis.‘Dispersal limitation’ in this faunally intact Amazonian forest is primarily attributable to a scant rain of dispersed seeds, i.e. fecundity limitation, whereas the distribution of dispersed seeds, being random for most species, appears adequate. Evidence from this and earlier research at the same site indicates that the per-capita success of dispersed seeds is many times higher than that of undispersed seeds. Thus, seed dispersal kernels that do not distinguish between dispersed and undispersed seeds are likely to be biologically misleading.