452 Bulletin American Museum of Natural History. [Vol. XX"\^II, 



the posterior hook or toe of the magnum becomes part of the articular sur- 

 face for digit III, (c) the broad overlapping of the metacarpals, etc. But 

 all these and other peculiarities present nothing inconsistent with derivation 

 from the carpus of Lower Eocene Perissodactyls (p. 398). 



Sirenia. The carpus in Manatus contains the principal elements, with 

 the exception of the centrale. The lunar rests mainly on the magnum but 

 retains the connection with the unciform. 



Cetacea. The carpus in Globiocephalus retains elements which are 

 interpreted by Flower as scaphoid, lunar, cuneiform, trapezium (?), trape- 

 zoid and unciform. The lunar-unciform connection is retained. 



Summary in regard to the carpus: The principal conclusions of the 

 preceding section are as follows: 



(1) The structural i)rototype of the mammalian carpus is realized in 

 the Permian and Triassic Therapsida in which the carpus differs from that 

 of mammals chiefly in retaining two elements (centrale 2, and distal carpale, 

 5) which afterward disappeared at least as separate elements. 



(2) The essential features of the interlocking type, namely the scaplio- 

 centrale-magnum and lunar-unciform contacts are present in many Ungui- 

 culate orders and are probably a primitive mammalian character. 



(3) In Unguiculates the retention of the grasping function, and of a 

 divergent pollex, favors the development of oblique facets and of a strongly 

 interlocking carpus (e. g., Didclphis, Talpa, Creodonts, Primates). 



(4) The loss of a divergent pollex and the development either of 

 ambulatory and cursorial habits or of great weight often favors the flattening 

 of the carpal facets in horizontal planes and the development of the serial 

 carpus {e. g., certain Insectivores, Rodents, Hyanodon, Hyrax, Proboscidea, 

 etc.). 



(5) The Perissodactyl and Artiodactyl manus have both been derived 

 from different varieties of the incipiently interlocking type. In both also 

 the interlocking features became emphasized, the serial features more or 

 less suppressed. 



(6) The observed ami inferred modes of evolution of the carpus in 

 mammals suggests first the complexity of the factors that have contributed to 

 the results and secondly the apparent inadequacy of explanations which 

 take into account only natural selection on the one hand and adaptive fitness 

 on the other. For exami)le, it might be assumed that the rectangular char- 

 acter of the carpals in the Proboscidea was largely an adaptation to great 

 weight, but since the carpals are almost equally rectangular in the small- 

 bodied Hyrax, it is quite possible that in the Proboscidea this condition was 

 established before great weight had been acquired. 



(7) In the different orders different elements of the carpus seem to be 



