393 
(a¢) is less deep in proportion to its breadth; its lower border has not the pair of low 
tubercles (fig. 5, 4, t); the hemal surface of the hypapophysis is produced downward 
and backward into the guasi-hemal spine, hy (Mivart, figs. 2-7); this is not present 
in Dinornis, but is replaced by a pair of low tuberosities, fig, 2, hy (which productions 
served for the attachment of the * longus colli’), as in Apterya, but are variable. The 
difference between Dinornis and Struthio in the relative size of the vertebrarterial 
foramina v is well marked ; the larger size of the canal in Dinornis relates to its better- 
developed brain: the roof of the neural canal is relatively less extended from before 
backward in Struthio; it is convex, rough or irregular in surface, with a feeble indi- 
cation of a medial ridge at the fore part in Dinornis (fig. 1, ), and with a hyper- 
apophysis (ib, 4p) as a low tuberosity above cach postzygapophysis. 
The postaxial articular surface presents, in Dinornis, a subquadrate convexity, and is 
not flat transversely in the present species or individual: the upper shortest border is 
moderately concave ; the lower longest border is framed, as it were, by a backwardly 
extended ridge, of which the pair of tubercles (fig. 5, ¢, ¢) form part, The postzyga- 
pophysial facets (2’, 2’, fig. 2) look more obliquely backward than in Sfruthio, where 
their aspect is almost wholly inward or ‘ mediad,’ 
In the general though slight convexity of the postaxial articular surface of the 
atlantal hypapophysis, in the slenderness of the pardiapophysial bar (fig. 3, pd), 
defining outwardly the yertebrarterial canal, and in the parial disposition of the 
hypapophysial tubercles, the atlas of Dinornis elephantopus? in the main agrees with 
that of D. rebustus and D, maximus. 
In Struthio the neural arch has a less relative antero-posterior breadth, and the 
same proportional difference prevails in the guasi-centrum ; the processes, ¢, ¢, in fig. 3, 
are not developed; the preaxial cup has a wider upper emargination. 
The length of the axis in Dinornis mawimus (fig. 4) is about four times that of the 
atlas, and equals about 24 inches. The preaxial surface of the centrum (fig. 5, cit) is 
twice as broad as high, and is concave transyersely, but less deeply than it is meuro- 
hemally. It is not prolonged neurally, as in the Ostrich, “ on to each side of the 
base of the odontoid process ”*; a slight non-articular concavity separates it on each side 
from the convex articular surface on the under part of the odontoid (figs. 4 and 7, 
ca, 0). This surface is convex, narrower and more produced than in the Ostrich‘. 
The postaxial surface (figs. 4 and 6, pc) has reverse proportions to the preaxial one, 
the longest diameter being vertical. It is divided into a pair of narrow vertically 
concaye facets by a still narrower medial tract, the transverse contour being thus rather 
angular than, as in Struthio and most birds, convex. The under border is nearly straight, 
and the transverse extent of the neural margin slightly exceeds that of the hwmal one. 
' P49, Pl, XIII, fig. 2 and Pl. XIV, a. 
2 PF. 232. 
* Mivart, Joc, cit, p, 301. 4 Ibid. fig. 12, 0. 
3 Q 



