Direct affinity of Endothiodon (ἐνδοθί, within) with Placodus is unlikely; the same applies to the Dicynodontia, although the restriction of the teeth to the palate seems to point as much to the former genus as do the toothless cutting edges of the jaws to the forms like Oudenodon.

Other Theriodont reptiles have been described from the upper Permian of Russia, for instance Deuterosaurus and Brithopus, but the determination rests upon insufficient fragments. North America has yielded many strange Theromorphous fossils, some of which may belong to the Theriodont order, while others seem to be intermediate between this and the other orders. Diadectes of Texas, for instance, seems to be a Theriodont creature; while in Empedias molaris, with a skull about 8 inches in length, the teeth form an uninterrupted series without distinct canine tusks, and the incisors are distinguished from the molars only by the transversely broadened shape of the latter. Very small teeth are arranged along the median line of the vomer and united palatine bones. In Clepsydrops, Dimetrodon, and Naosaurus of Texas the teeth are differentiated into incisors, canines, and molars, although not so regularly as in the typical Theriodont forms described above, one or more pairs of teeth being enlarged into canine-like tusks. In the latter two genera the spinous processes of the thoracic vertebrae are enormously elongated, standing up vertically to a height of 2 feet, while the centra of the vertebrae measure only one inch in diameter. In Naosaurus claviger these upright spines carry on either side half a dozen transverse projections. Stereorhachis of the Permian of France is typically Theriodont in the structure of its shoulder-girdle, humerus, and pelvis, but the dentition is composed of 3/3 incisors, no canines, and 6/10 pointed molars.

The following genera have been placed by Seeley in the family Gomphognathidae. Microgomphodon, with broader and less prominently multicuspid teeth than those of the typical Theriodonts, seems to lead to Gomphognathus, which has the following dentition: i. 3/3, c. 1/1, m. 12/12, with a long diastema between the canines and molars, some of which latter are nearly as broad as they are long, and have comparatively low tubercles on the crowns. The skull is remarkably like that of a Carnivorous Mammal. There are incisive foramina behind the premaxilla. The maxillaries and palatines form a united palatal roof, and behind them open the choanae. The occipital condyle is kidney-shaped. The mandible is most extraordinary, approaching that of the Mammalian, especially the Marsupial type, except that it is still composed of several pieces. The articular facet for the mandible is borne by an outward or lateral projection, while the bulk of the posterior half of the jaw projects inwards like a broad flange, undoubtedly recalling the so-called inner inverted angle of the Marsupial jaw. The coronoid process is large and extends far into the temporal fossa. Nearly the whole skeleton of Microgomphodon is known; the lumbar ribs are broadened and overlap as in Cynognathus, and the mandible is typically compound, so that there is no doubt about the affinities of this genus with the Theriodontia. It throws light upon Gomphognathus and the three likewise South African genera Diademodon, Trirachiodon and Tritylodon, which are all known from imperfect skulls only. Their teeth are restricted to the jaws, the molars have flat, multitubercular crowns and bear an extraordinary resemblance to those of Mammals. Some of the molars of Tritylodon are said even to possess two roots, but this point, absolutely unique in Reptiles, but common in Mammals, is not certain. The few upper incisors of Tritylodon are rather large, chisel-shaped, and extend like those of the Rodent-type back into the maxillaries; canines are absent, leaving a diastema. Trirachiodon has prominent canines, the five upper molars are multitubercular, rather flat, and much broader transversely than in the longitudinal direction. Still, even these creatures, with skulls of the size of that of a small fox, possessed distinct prefrontal and postfrontal bones, and are, at least in this respect, typical Reptiles.

Order III. ANOMODONTIA.

The cranium is not roofed in. The pedicle for the suspension of the lower jaw is much elongated, slants slightly forwards, and is composed of the long quadrate, which is laterally overgrown by the squamosal bone. The teeth are restricted to a pair of strong, tusk-like canines, or they are altogether absent. The margins of the upper and especially those of the lower jaw are trenchant, and were possibly furnished with a thick horny armature like those of tortoises.

Dicynodon, with many species from the Karroo formation of South Africa, reached formidable dimensions. The thick, curved skull is in size and outline not unlike that of a large lion, hence D. leoniceps, D. tigriceps, etc. The zygomatic arch is almost mammalian, except that the posterior boundary of the orbit is formed by a distinct postfrontal bone. The nostrils are lateral. The canine tusks (Fig. 54, E, p. 280) are very large. The choanae open behind the rhomboid vomer and between the separated palatine bones, which are posteriorly confluent with the medially united pterygoids. The latter send out flat extensions, along the lateral side of the palatines; these extensions reach the maxillaries and probably represent the ectopterygoids. The occipital condyle is distinctly triple, being equally composed of the basi- and latero-occipital bones.

The three bones of the shoulder-girdle meet at the glenoid fossa; the scapula has the indication of a spine. The pelvis is stout, attached to four or five vertebrae, converting the latter into a very Mammalian-like sacrum, the position of which lies distinctly in front of the acetabulum. The latter is closed, composed by the three pelvic bones. The pubes and ischia are fused together, leaving only a very small obturator-foramen. The limbs are plantigrade and pentadactyle, very stout; the humerus and femur have enormous crests.

Oudenodon, of which several species have been described, is so much like Dicynodon, except for the complete absence of teeth, that it has been suggested that these skulls belong to females of this genus. This view is strengthened by the fact that tusk-like canines exist, or are absent in some of the species which have been described as Cistecephalus, a genus closely allied to Dicynodon. The latter, which, like Oudenodon and Cistecephalus, occurred in Africa, extended also into India, D. orientalis having been found in the Panchet formation of Bengal, of transitional age between the Permian and Triassic epochs. Oudenodon rugosus, on the other hand, has been described from the Ural.

Gordonia and Geikia, of the New Red Sandstone of Elgin, are known from their skulls only, but these are so well preserved that there is no doubt about their close relationship to the typical South African Dicynodontia. The skull of Gordonia is about 7 inches long and 4 inches high. The canines (Fig. 54, D, p. 280) are reduced to short, but thick, conical tusks. The most remarkable feature is the very elongated squamoso-jugal arch, which arises moreover from the dorsal end of the long squamoso-quadrate pedicle. The two wide and long temporal fossae are dorsally divided by narrow parietal crests. There is a distinct interparietal bone, and the usual interparietal foramen. The choanae are united and lie within the palatines, which themselves are united; the large lateral palatal foramina are otherwise enclosed by the pterygoids, quadrates, and laterally by the squamoso-jugal arch.

Order IV. PLACODONTIA.

These are the latest and last members of the Theromorpha, unfortunately known from skulls only, from the Muschelkalk or Middle Trias of Germany and Russia. The skull of Placodus gigas is about one foot long, rather high and triangular owing to the lateral expansion of the temporal arches, which diverge posteriorly. The squamoso-jugal arch is very broad, and most of the posterior border of the orbit is formed by the large postorbital bone. The maxillary bone seems to extend back to beyond the level of the orbits. The choanae lie behind the premaxillaries. The palatines and pterygoids are fused in the middle line, forming a broad bony palate, which, owing to the broad, posteriorly extended wings of the pterygoids, much resembles that of the crocodiles. The teeth are very remarkable. There are two or three stout, conical, or chisel-like teeth in each premaxillary bone, and three to five broad and flat maxillary teeth; three pairs of huge, broad, and quite flat teeth are crowded together and fill up the whole vomerine and palatine portion of the palate. These crushing teeth indicate that Placodus probably lived upon hard-shelled molluscs, and this would be in conformity with its occurrence in the Muschelkalk, which is a strictly marine deposit and full of shells. Another closely allied genus is Cyamodus, one species of which is known from Russia. The teeth are fewer in number and not so large as those of Placodus.

CHAPTER IX

CHELONIA–ATHECAE–THECOPHORA

Sub-Class IV.–CHELONIA.

There is no mistaking a tortoise. The shell and the horn-covered toothless jaws separate them from all other four-footed creatures.

They may be described as terrestrial or aquatic, pentadactyle reptiles, with walking limbs or with paddles; ribs with capitular portions only, two sacral vertebrae, humerus with entepicondylar foramen, pubes and ischia forming symphyses, quadrate bones fixed, jaws without teeth, but with cutting horny sheaths. Trunk encased in a bony shell, composed of numerous dorsal and ventral dermal bones, forming a carapace and a plastron, which may or may not be covered with horny shields. Copulatory organ unpaired, cloacal opening more longitudinal than round, never transverse. Oviparous.

It is customary to distinguish the marine, paddle-limbed kinds as Turtles, the others as Land- and Water-tortoises.

Tortoises occur already in the Trias. They reached their greatest development towards the end of the Mesozoic and in the earlier Tertiary periods. They are now comparatively reduced in the number of families and genera, although they are still represented by about 200 species. The sub-class as a whole is cosmopolitan, but does not occur in the colder regions.

Their origin is quite unknown. Of recent groups only the Crocodilia and the Rhynchocephalia come into consideration. Combination of these groups with the Chelonia leads to some unknown forms whence also the Theromorpha have arisen. Palaeontology does not help us, all the leading, main groups of Chelonia having been in existence in the earlier Mesozoic ages, and Palaeozoic Chelonia are still unknown. We can, however, to a certain extent, reconstruct an ideal primordial Chelonian by assigning to it all the ancestral characters actually observed in recent and fossil kinds, and by reducing to simpler conditions those features which we know to be more or less exaggerated specialisations. It is reasonable to assume that originally each metamere, except those of the anterior half of the neck and the posterior half of the tail, carried a transverse series of dermal plates, covered with horny shields, while the trunk, according to the greater bulk of the body, increased in size, converging towards the root of the neck and tail. By concentration, reduction of the number, and increase in the size of some of the remaining plates and shields, the skull assumed its characteristic box-like shape, the neck and tail becoming at the same time free. Chelonia are without doubt descendants of terrestrial, or at least semi-aquatic reptiles, and the marine paddled forms subsequently developed from terrestrial kinds.

Classification of Chelonia.–After many vicissitudes it was recognised that the Chelonia cannot naturally be divided according to the modification of their feet. The Trionychoidea were clearly separated from the rest by Stannius in 1854. Cope, in 1870, was the first to emphasise the important character of the mode in which the neck is either bent sidewards (Pleurodira) or withdrawn in an S-shaped curve in a vertical plane (Cryptodira); and he also separated Sphargis as Athecae from all the other Chelonians, for which Dollo in 1886 proposed the term Thecophora. The division of the latter into recognisable families, based upon reliable, chiefly internal, skeletal, characters, has been effected by Boulenger;^[127] and his classification has been adopted in the present volume, after intercalation of the more important fossil forms. The relationships between these various families may perhaps be indicated as follows:–

The guiding taxonomic characters are fully mentioned at the head of the different families, and are mostly internal. The following "key," adapted from Boulenger, and based upon external characters, is preferable for practical purposes.

For the position and names of the horny shields see Fig. 61 on p. 315.

The vertebrae are, sometimes in the various regions of the same individual, amphi-, opistho- or pro-coelous, or even biconvex. Traces of the chorda remain longest in the middle of the centra. Intercentra occur regularly on the first two or three cervicals, and then again in the tail as paired or unpaired nodules, or as short chevrons. The latter occasionally fuse with the caudal end of their centra. Intercentral discs of fibrous cartilage occur regularly in the neck and tail. The ribs develop originally in the same transverse level with these discs, and frequently the anterior thoracic vertebrae retain this intercentral or intervertebral position throughout life. Farther back they often show a gradual change from the intercentral to a more central and ultimately remarkable to a purely neural attachment. In all the Chelonia the ribs are devoid of the tubercular portion.

The cervical vertebrae have no ribs, except mere traces in the shape of small nodules. On the tail the ribs are often large, and, when fused with their neural supports, look like transverse processes; the whole arrangement exactly resembles that of Crocodilia. The first pair of thoracic ribs, those borne by the ninth vertebra, are peculiar. They arise from the anterior portion of the centrum, are much reduced, sometimes to mere threads of bone, and lean against the anterior rim of the second pair of ribs, in many cases without reaching the carapace. The next following ribs, those of the tenth to the sixteenth vertebra, are intimately involved in the formation of the first to seventh costal plates. The ribs of the two sacral vertebrae sometimes remain quite distinct throughout life, just touching the upper ends of the iliac bones; but since these find a much more effective support in the shell, the distal ends of the sacral vertebrae fuse with the eighth, or so-called last, pair of costal plates.

The neural arch of the ninth vertebra rests upon its centrum; but the neural arches of the other trunk-vertebrae, although long, rest upon two centra; retaining, like the ribs, their original intercentral position; and in most cases the neuro-central sutures remain throughout life. The atlas and the last cervical vertebra deserve special attention. In many tortoises, e.g. Trionyx, Clemmys, Testudo, the three constituent parts of the atlas, namely, the neural arch, the centrum, and the intercentrum or first pair of united basiventralia, do not ankylose, but remain loosely connected; and the first centrum, instead of forming an odontoid process, remains movably attached to the second centrum, although it sometimes carries, and fuses with, the second intercentral piece. In other tortoises, e.g. Platemys and Chelys, however, all the parts of the atlas co-ossify and form a complete, solid vertebra which articulates by a concavo-convex joint with the centrum of the second vertebra. The normal number of cervical vertebrae is eight in all Chelonians. The first spinal nerve issues between occiput and atlas, all the others behind the neural arches of their vertebrae. The last, or eighth cervical, owing to the retractility of the neck, forms elaborate joints; its centre fits with a knob into a cup of the ninth, and its post-zygapophyses form broad, curved articulating concave facets for the reception of the anterior zygapophyses of the fixed ninth vertebra. In the Trionychidae the zygapophyses are most elaborate, and they alone articulate with the ninth vertebra, while the centra do not join, but remain, or rather become, separated by partial resorption. In the Chelonidae, in conformity with the non-retractile and short neck, all the cervical joints are much reduced.

The skull (cf. Fig. 54, H, I, K, p. 280) agrees fundamentally with that of Sphenodon and of the Crocodilia, but it is characterised by several special features. There are no ectopterygoids or ossa transversa; no lacrymal bones, no interparietal or pineal foramen; the vomer is unpaired and the nasal bones are mostly absent, unless they are fused with the prefrontals. The premaxillae are very small. The single vomer forms a septum between the choanae; and these are, except in Sphargis, ventrally roofed over by wings sent out by the palatines. The latter form a continuous bony roof to the mouth with the pterygoids, and these diverge posteriorly, being connected suturally with the quadrates, lateral and basi-occipital bones, and with the unpaired basi-sphenoid, which appears between the basi-occipital and the diverging pterygoids, but is in most cases to a great extent overlapped by the latter. The occipital condyle is distinctly triple; the basi-occipital sometimes helps to border the foramen magnum. The supra-occipital sends out a long vertical blade, directed backwards and generally projecting far over the neck, for the attachment of the powerful cranio-cervical muscles. The quadrate is very peculiar. Firmly attached, and hemmed in on nearly all sides by the neighbouring bones, it stands nearly vertically and forms a broad articulating surface for the mandible. Its posterior side shows either a transverse, horizontal groove, in which lies the columella auris, or the groove is transformed into a more or less closed canal. Moreover, the hinder lateral margin of the quadrate forms most of the tympanic frame; its margins being curved backwards, leaving in the Cryptodira, however, a wide notch behind; in the Pleurodira this part of the quadrate is transformed into a trumpet, the wide rim of which, forming a complete ring, carries the tympanic membrane. The tympanic cavity thus formed often leads into a deep recess which extends beneath the squamosal towards the opisthotic and bears some resemblance to the intricate tympanic recesses which pervade that region of the Crocodilian skull.

Dorsally the quadrate is broadly overlaid by the squamosal, which frequently forms an arch with the parietal. Anteriorly the quadrate is connected through a variably sized quadrato-jugal with the jugal; and this, by joining the maxilla and postfrontal, helps normally to form the posterior rim of the orbit. All the bones which border the temporal fossa vary much in extent in the different groups of Chelonia. The extremes are represented by Cistudo and Geoemyda, in which the bony infratemporal arch is absent, owing to the loss of the quadrato-jugal; and on the other hand by the Chelonidae and by Sphargis, in which the whole temporal region is covered over by an additional "false cranial" roof. This roof is produced chiefly by lateral wing-like expansions of the parietal and postfrontal bones, which meet the likewise much expanded jugal, quadrato-jugal, and squamosal bones. In the lower diagram of Fig. 63 (Chelone mydas) the squamosal has been removed, and the other bones have been reduced to their normal, or rather primitive condition, for comparison with the external view of the complete skull of the same animal. The lower diagram shows also the connexion of the pterygoid with a descending process of the parietal; this column, paired of course, usually contains a separate bone, the epipterygoid, the portion between Ptg and Par.

The hyoidean apparatus is well developed, and sometimes assumes large dimensions, especially in Chelys. The two pairs of "horns" are the first and second branchial arches, whilst the hyoid arches are reduced to a pair of small, frequently only cartilaginous, nodules attached near the anterior corners of the basis linguae, which generally fuses with the os entoglossum in the tip of the tongue.

The pectoral arch consists of a pair of long coracoids sloping obliquely backwards, the distal cartilages of which scarcely touch each other in the middle line, and the scapulae. The upper end of the scapula frequently touches the inside of the first costal plate, protected by a cartilaginous pad. Near the glenoid cavity arises a long process (PC in Fig. 65), placed transversely and approaching its fellow. The distal end is connected with that of the coracoid by a fibro-cartilaginous band. The homology of this scapular process is not quite clear. The band just mentioned favours the idea that the process represents the precoracoid, but its being an outgrowth from the scapula suggests that it is merely the much enlarged acromion. It certainly does not represent the clavicle, which forms part of the plastron: and this is not in contact with the shoulder-girdle at all.

The pelvis is strong. Ilium, pubis, and ischium meet at the acetabulum. The dorsal end of the ilium is generally broadened, and is attached to one or both sacral vertebrae, but it is also in contact with the superimposed last costal plate. This additional connexion often becomes predominant and the sacral vertebrae are partly or completely relieved of the iliac support, fusing in this case more or less with the costal plates. The pubes have strong lateral processes, directed obliquely forwards and downwards. The pubes and the ischia, which latter are much smaller, form broad symphyses, and these are connected with each other by a longitudinal cartilaginous band (Chelone, Trionyx); or the connecting bridge is broad and quite ossified (Testudo), forming in the latter case two roundish obturator-foramina. Cartilage frequently remains at the anterior end of the pubic symphysis, and a smaller, longer, and narrow piece of cartilage extends sometimes backwards from the ischiadic symphysis, as the so-called hypo-ischium. In the Pleurodira the ends of the ilia, and those of the lateral processes of the pubes, are much broadened and firmly ankylosed with the posterior costal plates and with the xiphiplastron respectively.

The limbs are typically pentadactyle and complete, and are most primitive in water-tortoises, e.g. Chelydra and Emys, in which the carpus consists of the typical ten separate elements, including the pisiform. In Testudo the centrale is fused with the intermedium, and the first three distal carpals are also fused together. In the marine turtles the limbs are transformed into paddles, but all the bones retain their independence; the pisiform and the first metacarpal are enlarged and flattened, thereby giving additional width to the paddle. The tarsus remains less primitive; the centrale and the proximal elements have a tendency to fuse together, most completely in land-tortoises; the fifth distal carpal is enlarged, and stands out hook-like from the rest. The number of the phalanges of the fingers and toes varies slightly. It is noteworthy that none of the Chelonia possess more than three phalanges. The three middle fingers and toes have mostly three phalanges; the pollex and hallux have always two; the number of phalanges of the fifth finger varies from three to one, of the fifth toe from two to none. The greatest reduction occurs in Testudo and its allied genera of typical land-tortoises, Homopus, Pyxis, and Cinixys, the formula for the fingers being 2, 2, 2, 2, 2 or 1, and 2, 2, 2, 2, 0 for the toes. In Pelomedusa all the fingers possess two phalanges only, owing to fusion of the first and second phalanges with each other.

The shell, which is the most characteristic feature of the Chelonia, consists of the dorsal "carapace" and the ventral "plastron." Each is composed of a considerable number of bony plates which arise as ossifications of nearly the whole thickness of the cutis, only a thin layer of subcutaneous connective tissue remaining soft and lining the inside of the shell. We restrict ourselves to a description of the shell of the Thecophora, leaving the discussion of the peculiar shell of Sphargis to p. 336 f. Very young tortoises are still soft, and the plates which are beginning to ossify are not yet suturally united. The plastron (Figs. 66 and 67) consists of the paired epi-, hyo-, hypo-, and xiphi-plastral plates, and the unpaired endo-plastral plate.

The latter is homologous with the interclavicle, the epi-plastra are homologous with the clavicles of other Reptiles, while the other pieces are genetically derived from, and are further modifications of, the so-called abdominal ribs of the Crocodilia and Prosauria, These plastral plates are never in direct contact with the shoulder-girdle or with any other parts of the internal skeleton. In the young of all tortoises, and in the adult of the Chelonidae and Trionychidae, the several plastral plates enclose large, irregularly-shaped fontanelles. These are more or less filled up in the other groups; and in the Testudinidae especially the whole plastron forms one continuous mass. The navel is situated between the hyo- and hypo-plastrals. Both these pairs are broader than the others, and are connected with the carapace by means of several marginals. The connecting region is called the bridge. In several tortoises, e.g. Emys, the connexion with the marginals is formed by ligaments only and remains movable. In others, transverse, more or less perfect hinges are formed across the plastron. A rather imperfect joint between the hypo- and xiphi-plastrals develops with age in Testudo ibera. In Cistudo and Cyclemys a very effective hinge lies below the hyo- and hypo-plastrals, just in front of the bridge; and the anterior and posterior lobes of the plastron can be closed against the inner rim of the box, fitting tightly in Cistudo. In Pyxis the front lobe only is movable.

The carapace is composed of one median series, a right and left lateral series of costal plates, and a series of marginals which surround the whole. The median series consists of one large nuchal plate, normally eight neurals and one to three supracaudal plates. The characteristic feature of the neural plates is that they are firmly fused with the broadened neural spinous processes of the underlying vertebrae. The nuchal plate lies in front of the first thoracic or ninth vertebra; it overlies the last cervical vertebrae, with the eighth of which it is connected by ligament only; but the posterior corner of the plate often fuses with the spine of the ninth vertebra. In the Chelydridae, and still more in the Trionychidae, the nuchal sends out a pair of long rib-like processes, which either extend to below some of the neighbouring marginals, or their ends overlap those of the ribs of the second thoracic vertebra (e.g. Trionyx), or, lastly, they are in turn overlapped by the first costal plates (e.g. Cyclanorbis). Such rib-like processes are also present, well developed in the young, shorter in the adult, in the Dermatemydidae and Cinosternidae. It is possible that the nuchal plate represents the fused neural of the eighth and the costal plates of the ninth vertebrae. An indication of the compound nature of the nuchal may be found in the fact that two nuchals have been described in Chelydropsis carinata, a Miocene relation of Chelydra. Somewhat similar modifications have taken place in the post-sacral region. The one to three supracaudal plates are, namely, neurals which have lost their connexion with, or perhaps have never been fused with, the spinous processes of the movable tail-vertebrae. The number of neural plates is mostly eight, but there are sometimes individually nine or ten, the gradual suppression taking place first in the sacral region. When such a plate is suppressed the neighbouring costal plates usually close up and meet in the median line. In Cistudo, for instance, there are only seven normal neurals, the eighth pair of costals meet, and the original eighth neural is transformed into a supracaudal. In Cinosternum the sixth to eighth costals meet, separating the one supracaudal widely from the remaining five neurals. The meeting of the last pair of costals, with co-ordinate reduction of the neurals to seven, is almost universal in the Pleurodira; and this tendency is carried out to an extreme in the Brazilian Platemys and in the Australian Chelodina and its allies, in which all the costals meet in the middle line, and the neurals are completely suppressed. Every stage intermediate between complete neurals (Sternothaerus) and interrupted, vestigial, and vanished neurals, is still represented by some genus. This process takes place independently, both in America and in Australia, and is one of the most recently introduced modifications.

The costal plates arise, like the neurals, independently in the cutis, but they soon come into contact with the underlying cartilage of the ribs, which are long enough to reach the marginals. The ribs flatten, become surrounded by the growing membrane-bone of the plates, and the cartilage of the ribs, instead of ossifying, undergoes a process of calcification. Ultimately this is more or less absorbed, its place is taken by the dermal bone, which forms so to speak a cast of the rib, preserving in many cases the shape of the vanished rib, only the capitular portions of which remain unaffected. The number of costal plates is very constant, namely eight on each side, but some fossils have nine or ten, and there are still individual variations in recent forms, indicative of that number. In a large Chrysemys concinna I find the last pair of costals clearly composed of at least two pairs, and this same specimen has nine distinct neural plates.

The marginal plates are originally paired, almost always eleven pairs, very rarely ten or twelve; an unpaired posterior plate, the pygal, is always present, and is probably the result of fusion. In the Chelonidae large fenestrae remain between the costal and marginal plates, only covered by leathery unossified cutis, and of course by the horny shields. In the Indian fresh-water genus Batagur similar windows are gradually filled up with age, and the horny shields become extremely thin and almost confluent. On the other hand, in Testudo polyphemus, the bony shell, always very thin, becomes still thinner with age and finally fenestrated by absorption.

Great reduction has taken place in the carapace of the Trionychidae. The American species of Trionyx have only seven pairs of costal plates; in Cyclanorbis the neurals are reduced to two. The whole dorsal shell is much smaller than the body, and marginal plates are absent or merely vestigial. It is doubtful if the ossifications in the posterior half of the marginal flap of some genera are homologous with true marginals.

Externally the whole shell is covered, except in the Trionychidae, in Sphargis and Carettochelys with horny, epidermal shields. These are phylogenetically older than the dermal plates, and they do not correspond with them either in numbers or in position, although there exists a general resemblance in their arrangement. On the plastron we distinguish an unpaired or paired gular, and a pair of gular, humeral, pectoral, abdominal, femoral, and anal shields (Fig. 66). Sometimes there are also intergulars, paired in Macroclemmys and Chelys, unpaired in Chelone; in many of the Pleurodira an unpaired intergular lies behind the gulars.

The carapace of most Chelonians is covered with five neural, four pairs of costal and twelve pairs of marginal shields, the last of which often forms an unpaired pygal. In front of the first neural lies the nuchal shield, very variable in size, often absent. The Chelydridae, Dermatemydidae, Platysternidae, and Cinosternidae possess moreover several inframarginals, intercalated on the bridge between the marginal and some of the plastral shields. In many of the other families these inframarginals are restricted to the anterior and posterior corners of the bridge, as the so-called axillaries and inguinals, mostly small and variable. Lastly, Macroclemmys has several small supramarginals.

There are consequently eleven longitudinal rows of shields in all; by elimination of the supra- and infra-marginals they are reduced to seven rows. It is absolutely certain that the number of transverse rows also was originally much greater than it is now. The mode of reduction of the number of the neural and costal shields has been studied in Thalassochelys caretta (cf. p. 388.) The accompanying illustration (Fig. 69) shows some of the main stages actually observed in the reduction of these shields. The chief point is that certain shields are squeezed out, or suppressed by their enlarging neighbours. The ultimate result is the formation of fewer, but larger shields.

Each shield grows individually as follows. Every year, or rather during every periodically recurring period of growth, the area of the Malpighian layer belonging to each shield increases peripherally in size, and at the same time produces a new layer of horn. The original little shield, with which the tortoise is born, remains for years, often throughout life, as the so-called "areola;" it increases in thickness owing to the new layer of horn added from below, and peripherally the increase in size is indicated by the overlapping concentric rings. Each ring represents a year's growth, at least in tortoises which live in temperate zones, where hibernation means a complete suspension of growth. It is not known if the same applies to tropical species, which grow either throughout the year, or which undergo one or more periods of rest. The areola does not remain central; the growth is uneven. With age the oldest layers of the areola are frequently rubbed off, and the areola then appears enlarged.