Manipulation of bee behavior by inflorescence architecture and its consequences for plant mating.

Angiosperms display flowers in many three-dimensional arrangements, but the functional significance of this diversity is largely unknown. We examined influences of inflorescence architecture on pollination and mating by quantifying the responses of bumblebees to three architectures and then using these observations as the basis of a model that simulated pollen dispersal. On artificial panicles, racemes, and umbels, each with 12 identical flowers, bees visited one more flower, on average, on umbels than on panicles (with racemes being intermediate). In contrast to this weak response, the consistency of foraging paths among flowers differed strongly among architectures (raceme > panicle > umbel). The simulation model revealed limited differences in self-pollination and pollen export among the three inflorescence designs when all flowers presented and received pollen, because mating differences depended on only the number of flowers visited. In contrast, in simulations of inflorescences on which pollen receipt and presentation were segregated so as to minimize interference among flowers, the consistency of movement paths governed mating. In this case, racemes self-pollinated much less than umbels (with panicles being intermediate), and racemes exported much more pollen than umbels and panicles. These effects have diverse consequences for the evolution of inflorescence architecture, flower design, and sexual segregation.