[Footnote 4:  Such, at least, is the conclusion suggested by the proportions of the skeleton figured by Cuvier and De Blainville; but perhaps something between a Horse and an Agouti would be nearest the mark.]

It would be hazardous to say that Plagiolophus is the exact radical form of the Equine quadrupeds; but I do not think there can be any reasonable doubt that the latter animals have resulted from the modification of some quadruped similar to Plagiolophus.

We have thus arrived at the Middle Eocene formation, and yet have traced back the Horses only to a three-toed stock; but these three-toed forms, no less than the Equine quadrupeds themselves, present rudiments of the two other toes which appertain to what I have termed the “average” quadruped.  If the expectation raised by the splints of the Horses that, in some ancestor of the Horses, these splints would be found to be complete digits, has been verified, we are furnished with very strong reasons for looking for a no less complete verification of the expectation that the three-toed Plagiolophus-like “avus” of the horse must have had a five-toed “atavus” at some earlier period.

No such five-toed “atavus,” however, has yet made its appearance among the few middle and older Eocene Mammalia which are known.

Another series of closely affiliated forms, though the evidence they afford is perhaps less complete than that of the Equine series, is presented to us by the Dichobune of the Eocene epoch, the Cainotherium of the Miocene, and the Tragulidoe, or so-called “Musk-deer,” of the present day.

The Tragulidoe; have no incisors in the upper jaw, and only six grinding-teeth on each side of each jaw; while the canine is moved up to the outer incisor, and there is a diastema in the lower jaw.  There are four complete toes on the hind foot, but the middle metatarsals usually become, sooner or later, ankylosed into a cannon bone.  The navicular and the cuboid unite, and the distal end of the fibula is ankylosed with the tibia.

In Cainotherium and Dichobune the upper incisors are fully developed.  There are seven grinders; the teeth form a continuous series without a diastema.  The metatarsals, the navicular and cuboid, and the distal end of the fibula, remain free.  In the Cainotherium, also, the second metacarpal is developed, but is much shorter than the third, while the fifth is absent or rudimentary.  In this respect it resembles Anoplotherium secundarium.  This circumstance, and the peculiar pattern of the upper molars in Cainotherium, lead me to hesitate in considering it as the actual ancestor of the modern Tragulidoe.  If Dichobune has a fore-toed fore foot (though I am inclined to suspect that it resembles Cainotherium), it will be a better representative of the oldest forms of the Traguline series; but Dichobune occurs in the Middle Eocene, and is, in fact, the oldest known artiodactyle mammal.  Where, then, must we look for its five-toed ancestor?