• Premise of the study: In animal-pollinated plants, autonomous selfing may provide reproductive assurance when pollinators or reproductive partners are limited. Under such circumstances, the contribution of pollinator-mediated seed set to total seed production also may be more variable compared with situations in which pollinator abundances are high or populations consist of large numbers of individuals. Despite the widespread acceptance of the reproductive assurance hypothesis, only limited empirical evidence exists that autonomous selfing confers reproductive output and guarantees constant seed set under variable pollination environments.

• Methods: We performed emasculation experiments in 22 populations of the short-lived, monocarpic plant Centaurium erythraea in a fragmented dune landscape.

• Key results: Floral emasculations resulted in a significantly lower seed set compared with that of intact flowers. Seed set in emasculated flowers also declined significantly with decreasing population size and pollinator availability, whereas seed set of intact flowers did not depend on population size nor on pollinator availability. Variability in seed set among individuals was significantly lower in intact than in emasculated flowers and decreased significantly with increasing population size when flowers were emasculated but not in intact flowers.

• Conclusions: These results indicate that pollinator-mediated seed set is strongly dependent both on population size and on pollinator availability but that reproductive assurance through autonomous selfing guarantees relatively constant levels of total seed production, even when populations are small and/or pollinator limited. High variation in seed set of emasculated flowers suggests strong unpredictability in pollinator services in small populations.