OSTEOLOGY OF CARNIVOROUS DINOSAURS. 



97 



it reaches its maximum height on the side of the arch — a position that is maintained 

 throughout the rest of the series. This facet stands out prominently from the 

 side of the arch on a pedestaUike projection. The narrow spine of the third dorsal 

 is succeeded by a slightly wider one on the fourth, and it in turn by a much wider 

 platelike spine that apparently continues throughout the rest of the series. These 

 vertebrae show the zygosphene articulation found in Antrodemus and in the Sauro- 

 podous dinosaurs. 



Presacrals 21 and 22 both bear ribs, though the latter is within the anterior 

 wings of the articulated iha. Presacral 23, however, I am inclined to beheve, did 

 not have a rib, but was in contact with the ihum through its transverse processes. 

 The broad, flattened neural spine extends well above the superior border of the 

 ilium (see pi. 30) and was evidently firmly coossified with the succeeding spine 

 of sacral 1. Measiu-ements of the presacrals are given in the table below. 



Measurements of presacral vertebrae of Ceratosaurus nasicornis Marsh. No. 4~35, U.S.N.M. 



1 Estimated. 



2 Kim of centrum restored. 



Sacral vertehrae. — The lateral view of the sacral vertebrae is almost entirely 

 hidden by the attached iha; only the tops of the coalesced spinous processes and 

 the ventral portions of the centra show where they project above and below the 

 ilium, as sho^vn in plates 27 and 30. The ventral aspect, however, is beautifully 

 illustrated in figure 1 of plate 21 — an unpublished drawing made many years ago 

 imder the direction of Professor Marsh. 



There are five vertebrae in the complete sacral arcade. The centra of the 

 sacrals are very much reduced in size and all are firmly coossified, not only with 

 one another but with the iha as well, the fusion being so complete as to almost 

 obliterate the hne of suture. This great reduction in size of the sacral centra and 

 their close coalescence is one of the striking dissimilarities between Ceratosaurus 

 and Antrodemus, as may be clearly seen by comparing figure 1, plate 21, with figure 

 2, plate 9. 



The transverse processes are short, all excejjt those on S5 being supported by 

 two vertebrae. An examination of the specimen appears to indicate that their 

 outer ends do not meet, and m tlfis respect the drawing appears to be misleading. 



