Selection for a floral trait is not mediated by pollen receipt even though seed set in the population is pollen-limited

Seed production in natural populations is often limited by quantity of pollen received. This pollen limitation has the potential to generate natural selection through female function favouring certain floral traits (hereafter pollen-mediated' selection). Floral traits can, however, also be under selection via other mechanisms, and the influence of a particular trait on pollen limitation of individuals is rarely quantified. We took a trait-based approach to pollen limitation by adding a quantitative trait to a previously published model. The modified model predicted impacts of both pollen supplementation and pollen reduction in seed production when the trait influences pollen receipt, and when the trait influences other model parameters. We then manipulated pollen availability in a population of the alpine New Zealand herb Wahlenbergia albomarginata varying in petal colour to test a long-standing hypothesis that pale flower colour in that habitat is selected for via pollinators. Under pollen-mediated selection, supplementing pollination should eliminate fitness dependence on a trait, while reducing pollination could exacerbate it. In contrast, if selection through female function on a floral trait is not solely pollen-mediated, seed production should change with the trait value even when excess pollen is added. Furthermore, if the effect is on maximum seed production, the trait effect should increase with higher pollen availability as greater resources can be used for filling seeds from more fertilised ovules. In Wahlenbergia albomarginata, average seed production was strongly pollen-limited, with supplemental pollen increasing seeds by 43%. Selection through female function also favoured whiter flowers as quantified with reflectance spectrometry. This selection on flower colour was stronger when supplemental pollen was added, consistent with an effect of the trait on asymptotic seed production rather than an effect on pollen receipt. These results, along with previous behavioural studies of pollinators, show that pale flower colour, common in the habitat and often attributed to pollinator behaviour, is not maintained by pollen-mediated selection but instead through another mechanism. This study illustrates the value of combining experimental pollen supplements and reductions with trait variation to study selection of floral traits, especially in species where compatible pollen receipt is difficult to measure. Lay Summary