Selfing rates in natural populations of Echium vulgare: a combined empirical and model approach

1. We quantified geitonogamous selfing in Echium vulgare, a self-compatible , bumble-bee pollinated plant. A maximum estimate of selfing was determined using a paternity analysis with RAPDs. In the first experiment, bumble-bee s visited a sequence of virgin flowers. The percentage selfing increased ra pidly from 12% in the first flower visited, up to 50% in the 15th flower vi sited in the sequence. In the second experiment, when bees visited plants i n a natural population, the average selfing of plants increased with the nu mber of open flowers from 0% to maximally 33%. 2. The results obtained in both experiments are consistently lower than pre dicted by our model on pollen dynamics (Rademaker, de Jong & Klinkhamer 199 7). We modified the model on pollen dynamics to link it more to the field s ituation with observations on flower stage, flower opening and bumble-bee p reference, so that the bumble-bees encounter a variable number of pollen gr ains per flower. We also adjusted the parameters. If less pollen adheres to the bee (25% instead of 50%) after removal from the anthers, or if bees ar rive at a plant with more pollen grains (6000 instead of 4448), the predict ions of the model in regard to selfing could be improved but were still hig h compared with the observed selfing rates measured with RAPDs. 3. We suggest that the model is consistent with pollen dynamics in the fiel d. However, post-pollination processes like selective abortion could play a role in E. vulgare.