TESTUDINID. 369 
latter family. The ends to be subserved by this cony exity may be various. We may suppose 
that it would be much more difficult for a carnivorous animal to effect an entrance into such 
a shell than into one deprest and whose borders may be spanned by the j jaws of the enemy. 
Again, the vaulted shell provides greater space for the lungs; and the investigations credited 
by Hoffmann (Brown's Klass. and Ordn.: =a at P. 344) to Dr. Joseph Jones but really 
made by Louis Agassiz (Contrib. Nat. Hist. U. S., vol. , p- 283) show that the capacity of the 
lungs of Gopherus polyphemus is much greater a ra of Emydide having an equal length 
of hele except Terrapene, which has camels habits of life. Even the earlier known forms of 
the Testudinide, as Hadrianus and Stylemys, have the shell high and vaulted; and this fact 
suggests that the deprest form of some existing species may he due to adaptations to recent 
conditions. The flattened carapace of Gopherus polyphemus, for instance, may have relation 
to the burrowing habits of the species. 
Among the Emy didz the neural bones are usually hexagonal, with the broader end of the 
bone in fant: Such, too, is the case with some of the Testudinid#. In Stylemys nebrascensts 
we find this condition prevailing; but occasionally a specimen is found in which the second 
one is octagonal. In S. conspecta and S. capax, of the John Day beds, there is further advance- 
ment—the second is octagonal, while the third is tetragonal. Too much weight ought not to 
be attacht to single specimens having modifications of this sort; as is illustrated by the fact 
that in the American Museum of Natural History there is a specimen of Chelydra ser pentina 
in which the second, fourth, and fifth neurals are octagonal, while the third is tetragonal. 
Nevertheless, among the Testudinide a high degree of differentiation of the neurals is the rule. 
When we come to the genus TJ estudo itself we fad the alternation of octagonal with tetragonal 
neurals the prevailing arrangement. In Kinyxis, according to Boulenger’s figure (Cat. 
Chelonians, fig. 41), the neurals are hexagonal with the broad end behind, a form the reverse 
of that common among the Emydide. In Homopus the emydoid condition prevails, except 
that the third is tetragonal. 
The costals have suffered remarkable modifications among the Testudinida. In most 
other turtles these plates, from the second to the sixth inclusive, have the proximal and the 
distal ends of each costal of nearly the same width. In most species of Testudo and in Gopherus 
the proximal ends of the second, fourth, and sixth are much narrowed, while the distal ends 
are greatly widened. On the other hand, the proximal ends of the third and fifth are broad, 
while the distal ends are narrow. The proximal ends of costals 3 and 4 are in contact with 3 
neurals each. The 5 costals concerned appear thus to be dovetailed with one another. The 
mechanical advantages of this arrangement have been discust by A. Bienz (Rev. suisse 
Zool., etc., 111, 1895, p. 99). These modifications of the costals are not conspicuous in Had- 
rianus ane in Stylemys nebrascensis, but they become very manifest in the species of Stylemys 
from the John Day beds. In Testudo brontops of the lowest Oligocene, the neurals display a 
high grade of differentiation, and, in harmony with this, the upper ends of the costals are modi- 
fied so as to unite alternately with 1 and with 3 neurals; but there is no great difference in the 
widths of the upper and lower ends of each costal. 
In most turtles the position of the rib is very distinctly shown on the visceral side of each 
costal plate; and the heads of the ribs are broad and thick. These heads join their respective 
vertebral centra at the anterior ends of the latter, and at least the anterior ones come into 
contact also with the next centrum in advance. In the Testudinidz the ribs show only faintly 
on the under side of the costals and the rib-heads are usually greatly reduced. Ina specimen 
of Testudo radiata from Madagascar, having the bel 305 mm. long, the rib-heads of the 
second to the sixth costal plates may be 3o mm. long; but they are very slender blades 5 mm, 
thick horizontally and 2 mm. high, expanding sews toward the proximal ends. In 
Gopherus only the merest traces of the rib-heads remain of those belonging to the costals in 
front of the seventh; and these vestiges adhere to the centra and to the Aenea, In 7. tabulata 
the rib-heads are all much reduced. Many of these are wholly free from the costal plates and 
have become co-ossified with the under sides of the neural bones. Intermediate stages are to 
be seen in this species. This condition illustrates well the manner in which the ribs, probably 
free from the costals in the most primitive turtles, have become consolidated with them. 
When we examine Hadrianus we find that the rib-heads were somewhat reduced, but still 
of considerable size. In Stylemys they are still more reduced. A careful examination of a 
24, 
