On plants with many flowers, bumblebees tend to visit more flowers in sequence. This induces pollen transfer between flowers on the same plant (geitonogamy). Consequently, in self-compatible plants the selfing rate increases with the number of simultaneously open flowers on the plant, and pollen export to other plants in the population can be a decelerating function of the number of flowers. I argue that geitonogamy is important in relation to two phenomena. First, it may explain the low number of seeds per flower and the size-dependent sex allocation observed in some species. Applying sex allocation theory to the boraginaceous Cynoglossum officinale and Echium vulgare shows that hermaphroditism is stable in both species. The predicted evolutionarily stable values for seeds per flower are close to observed values in the field. The model generally predicts that seeds per flower increase with plant size. This prediction is fine for C. officinale but E. vulgare defies the theory. Second, geitonogamy is important in the evolution of dioecy, as already suggested by Darwin. With abiotic pollination the export of pollen to other plants is likely to be proportional to the number of flowers produced, while it decelerates in animal-pollinated plants. Dioecy can then evolve gradually in a species with abiotic pollination, provided that inbreeding depression exceeds 50%, geitonogamous selfing increases with the number of flowers and that genetic variation exists. With complete pollen discounting as in the models of geitonogamous pollination by animals, hermaphroditism is always stable. The model explains Darwin’s observation that dioecy is more common in wind-pollinated species and in species in which individuals are large, reproducing with many flowers.