[43] It may be said, with regard to this particular reflex, that it may perhaps be, so to speak, a mechanical accident, arising from the contiguity of the sensory and motor roots in the cord. But as this suggestion cannot apply to other reflexes presently to be adduced, it need not be considered. I will adduce but one further example in illustration of these general principles--passing at once from the foregoing case of comparative simplicity to one of extreme complexity. There is a well-known experiment on a brainless frog, which reveals a beautiful reflex mechanism in the animal, whereby the whole body is enabled continually to readjust its balance on a book (or any other plane surface), as this is slowly rotated on a horizontal axis. So long as the book is lying flat, the frog remains motionless; but as soon as the book is tilted a little, so that the frog is in danger of slipping off, all the four feet begin to crawl up the hill; and the steeper the hill becomes, the faster they crawl. When the book is vertical, the frog has reached the now horizontal back, and so on. Such being the facts, the question is--How can the complicated piece of machinery thus implied have been developed by natural selection? Obviously it cannot have been so by any of the parts concerned having been originally distributed among different individuals, and afterwards united in single individuals by survival (i.e. free intercrossing) of the fittest. In other words, the case is obviously one of co-adaptation, and not one of the blending of adaptations. Again, and no less obviously, it is impossible that the co-adaptation can have been _gradually developed_ by natural selection, because, in order to have been so, it must by hypothesis have been of some degree of use in every one of its stages; yet it plainly cannot have been until it had been fully perfected in all its astonishing complexity[44]. [44] Of course it will be observed that the question is not with regard to the development of all the nerves and muscles concerned in this particular process. It is as to the development of the co-ordinating centres, which thus so delicately respond to the special stimuli furnished by variations of angle to the horizon. And it is as inconceivable in this case of reflex action, as it is in almost every other case of reflex action, that the highly specialized machinery required for performing the adaptive function can ever have had its origin in the performance of any other function. Indeed, a noticeable peculiarity of reflex mechanisms as a class is the highly specialized character of the functions which their highly organized structures subserve.