Ignoring within‐flower self‐fertilization and inbreeding depression biases estimates of selection on floral traits in a perennial alpine herb

Abstract Within‐flower self‐pollination should be the major source of self‐fertilization in mixed‐mating species that present single or few flowers simultaneously. It is also an often unmeasured source of selfing in species with many flowers open simultaneously. In self‐compatible species in which pistil and stamen numbers vary, the rate of within‐flower selfing should depend on the number of pistils and stamens, the timing of flowering and the morphology of subsidiary floral traits. The intensity and direction of selection on these traits should thus also depend on the level of inbreeding depression. Here, we measured the dependence of the within‐flower selfing rate on floral sex allocation, phenology, petal length and floral stalk height in a population of the perennial herb Pulsatilla alpina (Ranunculaceae) in which most individuals had single flowers. We estimated inbreeding depression in the population by comparing inbreeding coefficients between parents and seed progeny using microsatellite markers. We then estimated selection on the measured traits via female reproductive success at the flower level and compared our estimates with a hypothetical scenario in which inbreeding depression was assumed to be absent. Inbreeding depression was estimated to be severe (0.95). The within‐flower selfing rate varied widely among flowers and depended positively on stamen number and negatively on pistil number and flowering date, supporting the predictions of a mass‐action model. The dependence of the selfing rate on the measured floral traits consistently predicted (non‐linear) patterns of selection under high inbreeding depression that were distinct from those under a hypothetical scenario of no inbreeding depression. Synthesis . While previous research has emphasized the importance of mass‐action mating on selfing among flowers of plants with large floral displays, our results demonstrate its importance for selfing within individual flowers. They also demonstrate the importance of accounting for both the selfing rate and inbreeding depression when inferring selection on floral and other traits via female fitness.