Pollinator behavior and deceptive pollination: learning process and floral evolution.

Some species of flowering plants engage in nonmodel deceptive pollination, attracting pollinators by large nonmimetic floral displays and providing no reward. Pollinators can learn to avoid deceptive plants and to favor nectariferous species. The reproductive success of these species is expected to be density dependent for two opposite reasons: the commoner cheating flowers are, the easier they are to avoid and the lower the quality of the patch, making it more difficult to recognize that unrewarding flowers are not profitable. When a deceptive species is made up of multiple floral variants, pollinators' learning could decrease the reproductive success of any particularly common floral variant. Within a population of deceptive plants, mean reproductive success could, therefore, vary with the number of floral variants. We investigate these three hypotheses by modeling the behavior of pollinators foraging in communities of deceptive and rewarding flowers. Simulations revealed that the reproductive success of deceptive flowers varies in a density-dependent manner and that floral variants can be submitted to negative frequency-dependent selection. We compare density dependence in nonmodel deceptive species to what is expected in Batesian mimics and discuss possible selection of morphological variants. Finally, we survey how pollinators' learning capacities can make mean reproductive success depend on morphological variability within a population.