As to the origin of the soft-shelled turtles there has been not a little difference of opinion. The earliest ones known in geological history date back only to about the middle of the Cretaceous; perhaps they branched off from the horny-shelled turtles somewhat earlier, but probably not much. There are some, however, who think that this group of turtles was very primitive, perhaps the most primitive, but the writer agrees with Dr. Hay in rejecting this view. Unlike those of all other turtles, the fourth digit in front and hind feet has one or two more phalanges than have other turtles. We have seen that the oldest known reptiles had the digital formula 2, 3, 4, 5, 3 or 4. Most other turtles have the same numbers of bones in the digits that mammals have, that is, two phalanges in the thumb and big toe and three in each of the other digits. The river turtles have a larger number in the fourth digit, either four or five. It seems to be a law that evolution is irreversible, and if so could the river turtles have been descended from forms with a less number of phalanges? But, the skeleton of the Trionychoidea resembles the more specialized turtles in so many ways that one can hardly believe they were all accidental or parallel. We may then assume that at about the time that the ordinary marsh turtles took to the sea to become marine, others took advantage of the fresh-water ponds and rivers, and in doing so, like the marine turtles, lost their horny epidermal shields, and became thinner in shape, thereby reducing the resistance to the water. Instead, however, of reducing the costal plates over the ribs, they retained them intact and complete for some reason or other, but lost instead the marginal row of bones, unlike the marine turtles which retained them even after they had lost nearly all of the costal plates. Possibly also they regained additional bones in the fourth digit, a sort of hyperphalangy like that of the more strictly aquatic reptiles. Or, possibly, they may have descended from some branch of the turtles which had not yet lost these bones, retaining them because they were still serviceable for swimming. We know nothing yet about the structure of the feet of the early turtles, and it is possible that not all had acquired the reduced phalangeal formula. In the development of aquatic habits the river turtles do not show the same degree of specialization in the limbs that the strictly marine forms do. The humerus (Fig. 131) is a slender bone, with the tuberosities for the attachment of the muscles situated near the proximal end. The radius and ulna are relatively short, and the foot is long. The hind legs, as would be supposed, are less highly specialized as swimming paddles, and are relatively smaller. Nevertheless the Trionychoidea present an interesting type of adaptation to water habits, both in body and in limbs. INDEX