The stigma of Gesneria pendulina sometimes protrudes far beyond, and is sometimes seated beneath the anthers; so it is with Oxalis acetosella and various other plants. I have also noticed an extraordinary amount of difference in the length of the pistil in cultivated varieties of Primula veris and vulgaris.
As most plants are at least occasionally cross-fertilised by the aid of insects, we may assume that this was the case with our supposed varying plant; but that it would have been beneficial to it to have been more regularly cross- fertilised. We should bear in mind how important an advantage it has been proved to be to many plants, though in different degrees and ways, to be cross- fertilised. It might well happen that our supposed species did not vary in function in the right manner, so as to become either dichogamous or completely self-sterile, or in structure so as to ensure cross-fertilisation. If it had thus varied, it would never have been rendered heterostyled, as this state would then have been superfluous. But the parent-species of our several existing heterostyled plants may have been, and probably were (judging from their present constitution) in some degree self-sterile; and this would have made regular cross-fertilisation still more desirable.
Now let us take a highly varying species with most or all of the anthers exserted in some individuals, and in others seated low down in the corolla; with the stigma also varying in position in like manner. Insects which visited such flowers would have different parts of their bodies dusted with pollen, and it would be a mere chance whether this were left on the stigma of the next flower which was visited. If all the anthers could have been placed on the same level in all the plants, then abundant pollen would have adhered to the same part of the body of the insects which frequented the flowers, and would afterwards have been deposited without loss on the stigma, if it likewise stood on the same unvarying level in all the flowers. But as the stamens and pistils are supposed to have already varied much in length and to be still varying, it might well happen that they could be reduced much more easily through natural selection into two sets of different lengths in different individuals, than all to the same length and level in all the individuals. We know from innumerable instances, in which the two sexes and the young of the same species differ, that there is no difficulty in two or more sets of individuals being formed which inherit different characters. In our particular case the law of compensation or balancement (which is admitted by many botanists) would tend to cause the pistil to be reduced in those individuals in which the stamens were greatly developed, and to be increased in length in those which had their stamens but little developed.
Now if in our varying species the longer stamens were to be nearly equalised in length in a considerable body of individuals, with the pistil more or less reduced; and in another body, the shorter stamens to be similarly equalised, with the pistil more or less increased in length, cross-fertilisation would be secured with little loss of pollen; and this change would be so highly beneficial to the species, that there is no difficulty in believing that it could be effected through natural selection. Our plant would then make a close approach in structure to a heterostyled dimorphic species; or to a trimorphic species, if the stamens were reduced to two lengths in the same flower in correspondence with that of the pistils in the other two forms. But we have not as yet even touched on the chief difficulty in understanding how heterostyled species could have originated. A completely self-sterile plant or a dichogamous one can fertilise and be fertilised by any other individual of the same species; whereas the essential character of a heterostyled plant is that an individual of one form cannot fully fertilise or be fertilised by an individual of the same form, but only by one belonging to another form.