Phenotypic Selection on Flower Color and Floral Display Size by Three Bee Species

Plants exhibit a wide array of floral forms and pollinators can act as agent of selection on floral traits. Two trends have emerged from recent reviews of pollinator-mediated selection in plants. First, pollinator-mediated selection on plant-level attractants such as floral display size is stronger than on flower-level attractant such as flower color. Second, when comparing plant species, distinct pollinators can exert different selection patterns on floral traits. In addition, many plant species are visited by a diverse array of pollinators but very few studies have examined selection by distinct pollinators. In the current study, we examined phenotypic selection on flower color and floral display size by three distinct bee species, the European honey bee, Apis mellifera, the common eastern bumble bee, Bombus impatiens, and the alfalfa leafcutting bee, Megachile rotundata, foraging on Medicago sativa. To estimate phenotypic selection by each bee species and for all bees combined simultaneously and on the same group of plants, we introduce a new method that combines pollinator visitation data to seed set and floral trait measurements data typical of phenotypic selection study. When comparing floral traits, all bee species selected on the number of racemes per stem and the number of stems per plant, two components of floral display size. However, only leafcutting bees selected on hue or flower color and only bumble bees selected on chroma or darkness of flowers. Selection on chroma occurred via correlational selection between chroma and number of open flowers per raceme and we examine how correlational selection may facilitate the evolution of flower color in plant populations. When comparing bee species, the three bee species exerted similar selection pattern on some floral traits but different patterns on other floral traits and differences in selection patterns were observed between flower-level and plant-level attractants. The trends detected were consistent with previous studies and we advocate the approach introduced here for future studies examining the impact of distinct pollinators on floral trait evolution.