The evolution of self-fertilization in hermaphrodites is opposed by costs that decrease the value of self progeny relative to that of outcross progeny. However, self-fertilization is common in plants; 20% are highly selfing and 33% are intermediate between selfing and outcrossing. Darwin proposed an adaptive benefit of self-pollination in providing reproductive assurance when outcrossing is impossible. Moreover, if outcross pollen receipt is inconsistent within or between years, these conditions likewise favour self-pollination, and this can result in a mixture of self and outcross seed production (mixed mating). Despite wide acceptance, the reproductive assurance hypothesis has lacked the support of complete empirical evidence to show that variable pollination can create both the ecological and genetic conditions favouring self-pollination. We recently showed in Collinsia verna that during periods of infrequent pollinator visits, autonomous self-pollination boosted seed output per flower, the key ecological condition. Here we show low inbreeding depression and marker-based estimates of selfing, demonstrating that when the pollination environment in wild populations necessitates reproductive assurance, selfing rates increase. We provide a complete demonstration of reproductive assurance under variable pollination environments and mechanistically link reproductive assurance to intermediate selfing rates through mixed mating.

Pollinator unpredictability favors evolutionary shifts from outcrossing to autonomous selfing, which provides reproductive assurance. Our goal was to quantify the reproductive assurance benefit of autonomous selfing and the stochastic nature of pollinator‐mediated pollen receipt using three wild populations of the annual species Collinsia verna (Scrophulariaceae) over three years. The timing of autonomous self‐pollination in C. verna ranges from competing to delayed, with more than half of the autonomous pollen arriving late in floral life. Floral density, identity of insect floral visitors, and visitation rate varied within and among years, with significant variation in pollinator community between years within populations. Pollinator failure reduced fruit set within and among populations across years. However, reproductive assurance selfing increased fruit set on average by 8% overall and up to 30% in one population. Despite this fitness benefit, fruit set (=Σfruits/Σflowers) estimated in pollinator‐exclusion cages was significantly less than the fruit set in paired open‐pollinated plots. In a series of experiments, we found no difference in fruit set from self‐ vs. outcross pollen, ruling out early inbreeding depression. However, flowers, hand‐outcrossed at the same time that autonomous selfing naturally occurs, produced significantly more fruits relative to flowers that received pollen only through autonomous selfing. These results suggest that pollen type (self/outcross) is less important to fruiting success than is the timing or efficiency of autonomous pollination relative to vector‐mediated pollination. Our results support the reproductive assurance hypothesis and suggest that autonomous selfing in C. verna is adaptive under variable pollinator conditions. Fruit set results from both pollinator‐delivered and autonomous self‐pollen, suggesting a “best of both worlds” scenario, with mixed mating a likely outcome.

Collinsia verna, blue-eyed Mary, has floral attributes of an outcrossing species, yet most flowers readily self-pollinate under greenhouse conditions. Here we describe the mechanism of self-pollination in C. verna via changes in relative positions of the stigma and anthers and late timing of receptivity, resulting in delayed selfing. Each flower contains four anthers that dehisce sequentially over ∼1 wk. Pollen that is not collected by pollinators accumulates in the keel petal and retains high viability (>80% pollen germination) up to the time of corolla abscission. The stigmatic surface does not become receptive until after the third anther dehisces. This overlap in the sexual phases is concurrent with a change in herkogamy during floral development. In most flowers (70%), the stigma has moved to the front of the keel and is positioned near the anthers when the third anther dehisces. Under field conditions, fruiting success of plants within pollinator exclosures was ∼75% of the fruiting success in open-pollinated plants (33% fruiting success via autogamy vs. 44% fruiting success, respectively). Collinsia verna plants in pollinator exclosures exhibit variation in autogamy rates within natural populations (range 0-80%). In addition, only half of naturally pollinated, receptive flowers examined had pollen tubes growing in their styles. In contrast, shortly after corolla abscission, nearly all flowers examined (96%) had pollen tubes in their styles. Thus we find that in C. verna, autogamy occurs late in floral development, which has the potential to provide substantial reproductive assurance, and that individuals vary in their ability to set fruit through this mechanism. We suggest that delayed selfing mechanisms may be overlooked in other species and that variable pollinator availability may play a significant role in the maintenance of mixed mating in species with delayed selfing, such as C. verna.

Previous reviews of plant outcrossing rate survey data have agreed that predominant selfing and predominant outcrossing are alternative stable states of mating system evolution. We reanalyzed the most recent data and plot outcrossing rates as a continuous variable rather than as a class variable. Wind‐pollinated species are indeed bimodal. However, the shape of the distributions for animal‐pollinated species reveals that intermediate rates of outcrossing are common (49% of species fall between 20% and 80% outcrossing). Consequently, we suggest that mating system is best considered a continuous rather than a discrete character of plant populations.