Abstract. In moving seeds from one location to another, dispersers define the ecological context that a plant will experience throughout its lifetime. Because seed deposition sites may benefit one plant life cycle stage but not others, dispersal may give rise to a conflict between a plant's dispersal needs (e.g., predator escape) and its requirements for successful establishment (e.g., suitable habitat). In this study, we assessed the effects of two components of seed dispersal by ants (seed burial depth and elaiosome removal) at several stages during the recruitment of Leucospermum truncatulum, an ant-dispersed plant from the fynbos shrublands of South Africa. A seed-predator exclusion experiment revealed that the removal of elaiosomes significantly reduced rodent predation on seeds buried 1 cm deep and increased the viability of seeds that were not consumed. In contrast, we found that elaiosomes had no effect on the probability of a living seed emerging as a seedling. In a seed burial depth experiment in which predators were not excluded, deeper burial significantly enhanced seed survival, but this benefit came at the expense of reduced seedling emergence, delayed emergence, and smaller cotyledon size. Despite the potential importance of these seedling characteristics, elaiosomes and burial depth had no significant effect on seedling survival through the first year of this study. In contrast, burial depth had strong positive effects on the fraction of seeds that persisted in the seed bank. Although dormant individuals may die before germinating and reproducing, those that persist at greater soil depths can play a crucial role in rescuing populations from local extinction in unpredictable environments. Given the extreme variability in fire frequency in fynbos, the net effect of seed dispersal by ants on plant populations will depend on how temporal fluctuations in fire regimes influence cost-benefit thresholds related to seed dormancy and seed mortality.