WHOLE- AND PART-FLOWER SELF-POLLINATION IN GLYCINE CLANDESTINA AND G. ARGYREA AND THE EVOLUTION OF AUTOGAMY.

The overall rate of self-fertilization can be viewed as the sum of two distinct processes: 1) self-pollination of all ovules in a flower (whole-flower self-pollination); and 2) self-pollination of some of the ovules in a flower, occurring together with outcrossing of the remaining ovules (part-flower self-pollination). In some situations these processes may be equated with different modes of self-pollination. A model of the mating system in which the progeny of separate fruits serve as the unit of observation is presented. The model partitions the overall rate of self-pollination into components attributable to whole- and part-flower selfing. When the mating system is estimated using information on marker genotypes from chasmogamous fruits in two species of Glycine together with the whole- and part-flower selfing model, the results indicate that the chasmogamous flowers in a subalpine population of G. clandestina underwent a significant level of whole-flower selfing, whereas in another, lower elevation population of G. clandestina and in a subtropical population of G. argyrea, they did not. This difference is thought to be related to the contrast in the variability of environmental conditions for insect-mediated pollination between the habitats sampled. In particular, the large component of whole-flower selfing observed in the subalpine population of G. clandestina may be due to self-pollination that is induced during periods unfavorable to insect-mediated pollination. It can be demonstrated that such induced selfing will be selected whenever environmental conditions are such that pollinator activity limits seed set, and moreover that induced selfing can result in the selection of overall levels of self-pollination that are intermediate between 0 and 1. Monte Carlo simulation is employed to show that ignoring the correlation of self-fertilization events that result from whole- and part-flower selfing may lead to biased estimates of mating system parameters.