230 Bulletin American Museum of Natural History. [Vol. X, 



same time. It appears moreover to the writer that the evidence 

 which has been brought forward to demonstrate such an excep- 

 tional condition is inadequate and is capable of different inter- 

 pretation, as follows : 



AmpJiiccelias having been shown to be in all probability rela- 

 ted to Camarasaunts, the only absolute basis of separation of 

 the remaining types is the number of sacral vertebrcc, as in the 

 definitions of Marsh and Cope : 



Apatosaiirus. Camarasaurus Atlantosaiirns. Broniosanrjis. 



3 sacrals. 4 sacrals. 4 sacrals. 5; sacrals. 



This definition does not separate Atlantosaiinis ixoxw Camarasau- 

 rus ; the type sacrum of Atlaiitosaurus (Fig. yi?) is, however, 

 fractured anteriorly, and this animal may therefore have pos- 

 sessed 5 sacrals, like 

 Brofitosaiirus (Fig. yC). 

 The three coalesced 

 sacral spines and the 

 single free sacral cen- 

 trum in our specimen 

 suggest the following 

 hypothesis, namely, 

 that Camarasaurus had 

 fire sacral vertebrce ; 

 three of these constantly 

 coalesced both by centra 

 and neural spines, two 

 others coalesced less con- 

 stantly atid possessed free 

 spines. 

 Such difference in the growth and degree of coalescence is shown 

 to be probable (i) hy jDiplodocus, in which at least three vertebrae 

 coalesce by centra but only two coalesce by spines ; (2) by our 

 specimen, No. 222, in which (closely resembling Brontosaurus in 

 every other respect) one centrum is free and three spines are 

 coalesced ; (3) by the type of B. excelsus, in which three spines 

 are coalesced, as figured. (See also Note 2, p. 233.) 



The removal to a distance of one or both of the freer sacral 

 vertebroe is quite explained by the fact that these skeletons are 



Fig. g. Afiatosaii) us ajax. Dorsal vertebra. After 

 Marsh, 1897, Plate xviii. Approximately yg natural size. 



