Darwin was right.
Just like any flowering plants, orchids produce thousands of pollen grains. But instead of releasing these through random patches to visiting pollinators, they offer them as a single package (err… hum, actually as two packages, but as a single gift).
Pollinia are specialised structures that are adapted for plant pollen dispersal. It does not act as a reward for pollination service: pollinators cannot eat them, since they are glued on their head (or sometimes on your proboscis if you happen to be a butterfly instead of a bee). The only aim of pollinia is to deliver pollen grains to the stigma of yet another flower, so that pollen actually fertilizes ovules and does not get eaten by starving pollinators.
There is a very striking feature of pollinia. They stand at the end of a stalk. When the other extremity is glued on a pollinator’s head, the stalk begins to bend. It curves more and more over time, until it is reaching a very specific position a few minutes later. This is actually placing the pollinia right into a position that fits the flower shape and allows for pollen deposition on the stigma.
If you’re a curious naturalist, just like Darwin was, you may ask why does it take the pollinia the time to bend? Why not being placed on the pollinator’s head so as to be ready to pollinate the next flowers already? Just because bending is not hundred per cent accurate and you therefore take the risk to fall short ahead and not to fertilize any ovule…
Darwin’s hypothesis was that the bending of pollinia is lasting as long as an average pollinator’s visit to the plant. In other words, the bending time would prevent self-fertilisation, since pollen would not be deposited on stigmas of the same inflorescence even if pollinators do visit several flowers.
The hypothesis was not properly tested until recently , and guess what?
… Darwin was right : bending is always longer than the average visit of pollinators to plants (the regression line is shifted ahead of the 1:1 line toward bending time).
This is right regardless of the orchid species (or even regardless of the two asclepiad species taken as controls), and regardless of plant-pollinator pairs: short visiting insects are those visiting orchids with short bending pollinia and orchids with long bending pollinia are those visited by insects that take a long time visiting the flowers… Amazing, isn’t it?
This leads me to the concluding remark. Creationists are usually not known for acknowledging their own mistakes, even when they are proven wrong. But even more: I’ve never read a single creationist to recognize that opponents are sometimes right. So was Darwin about orchids and pollinia. You don’t have to think evolution is true to acknowledge that Darwin was right, here, since the point is about bending preventing self-fertilization however the way those species came into existence…
Of course, it would not make much sense to be so finely tuned as to avoid selfing if it was not selected against… The issue in this case is first that bending can easily evolve through a microevolutionary process. Second, the process of selection is mechanistically easy to spot out: counter selection of plants that tend to self more because of bending too fast (while on the other hand, bending too slowly is also counter selected because it fails fertilization completely).
Anyway, as we all know, Darwin was right. So deeply right that it’s even more than a question of bending.
 – Peter C.I., and S.D. Johnson. 2006. Doing the twist: a test of Darwin’s cross-pollination hypothesis for pollination reconfiguration. Biology Letters 2 (1): 65-68.