Differential gender selection on floral size:: an experimental approach using Cistus salvifolius

1. From Darwin to the present day, sexual selection has been shown to be widespread in both animals and plants. Attractive floral traits in hermaphrodite plants evolve in response to selection acting simultaneously through male and female sexual functions, but knowledge about the relative strength of gender-specific selection is scarce. 2. We experimentally altered flower size of the hermaphrodite Cistus salvifolius in three sites differing in physical characteristics, and measured the effect of this manipulation on both male and female success. 3. More pollen was dispersed from the non-manipulated flowers compared with the size-reduced flowers. Standardized selection differentials for flower size through pollen dispersal were significantly positive in all three populations. By contrast, flower size had little effect on female success. Fruit set was always high and selection for this trait through fruiting was not significant. The number of seeds per fruit from bigger flowers was significantly higher than that from smaller flowers, but this difference was attributable exclusively to one population in which selection via seeds per fruit was significant. Flower size reduction had no effect on offspring quality. 4. Our results show that phenotypic selection through male function generally supports that flower size would evolve primarily through selection on male fertility (male function hypothesis). Selection through female function was apparent only in one population, as pollen limitation was absent or weak. We show the relative strength of selection through male and female functions depends on the pollinator context.