Pollen presentation theory (PPT) predicts that plant species typically pollinated by frequent and wasteful pollinators ought to be much more parsimonious and only gradually release pollen compared to plant species pollinated by infrequent pollinators that are efficient at delivering the pollen they remove. To test PPT, we compare the pollen presentation schedules and pollination systems in three related Epimedium species, having different pollinators. Results showed that differences in anther dehiscence and flowering traits resulted in different pollen packaging schedules. For E. sutchuenense and E. franchetii, a special 'roll-up' movement of the anther wall during anther dehiscence increased pollen removal compared to the dehiscence pattern in E. mikinorii, which lacked the 'roll-up' movement. Investigations revealed that honeybees had a higher pollen removal rate and lower stigmatic pollen load compared to bumblebees. In accordance with PPT, E. sutchuenense presents pollen sequentially and slowly for the frequent and wasteful honeybees. In comparison to E. sutchuenense, E. franchetii had a faster presentation rate and was adapted to the efficient and infrequent bumblebees. However, E. mikinorii was pollinated by both bumblebees and honeybees at high frequency and had the fastest pollen presentation. This pattern could reduce pollen wastage by honeybees and might be an adaptation to its short flower longevity (less than 1 day), to increase the chances of pollen deposition on stigmas. The study indicates that pollen presentation schedules can be a consequence of interactions among anther dehiscence, flowering traits and pollination environments for a given species.

Self-interference is one of the most important selective forces in shaping floral evolution. Herkogamy and dichogamy both can achieve reductions in the extent of self-interference, but they may have different roles in minimizing self-interference in a single species. We used four self-incompatible Epimedium species to explore the roles of herkogamy and dichogamy in avoiding self-interference and to test the hypothesis that herkogamy and dichogamy may be separated and become selected preferentially in the taxa. Two species (E. franchetii and E. mikinorii) expressed strong herkogamy and weak protogyny (adichogamy), whereas another two species (E. sutchuenense and E. leptorrhizum) expressed slight herkogamy and partial protandry. Field investigations indicated that there was no physical self-interference between male function and female function regarding pollen removal and pollen deposition in all species. Self-pollination (autonomous or facilitated) was greater in species with slight herkogamy than in those with strong herkogamy. Artificial pollination treatments revealed that self-pollination could reduce outcrossed female fertility in all species, and we found evidence that self-interference reduced seed set in E. sutchuenense and E. leptorrhizum in the field, but not in E. franchetii and E. mikinorii. These results indicate that well-developed herkogamy is more effective compared with dichogamy in avoiding self-interference in the four species. In genus Epimedium, herkogamy instead of dichogamy should be selected preferentially and evolved as an effective mechanism for avoiding self-interference and might not need to evolve linked with dichogamy.