246 2% SECTION — VERTÉBRÉS (SYSTÉMATIQUE) 
which are canines and not incisors. The grinding teeth have a striking 
similarity to those of the rhinoceroses, more particularly of Metamyno- 
don and Cadurcotherium, of the North American and European Oligo- 
cene respectively. So long as the feet remain unknown it will be impos- 
sible definitely to determine the taxonomic position of the Astrapotheria, 
but the more ancient members of this order, from formations older than 
the Santa-Cruz, show in their dentition such an approximation to the cha- 
racteristie molar-patterns of the other and specifically South American 
ungulate groups, as to indicate, in my opinion, a nearer relationship to 
the Toxodontia than to the Perissodactyla. It is a highly signifigant fact 
that in these more ancient genera the molar-patern is less rhinocerotic 
than it became in the Santa-Cruz epoch, a fact which, of itself, is sugges- 
tive of convergent development. 
Most remarkable of all the hoofed animals are the Läitopterna. In the 
Santa-Cruz beds we find the ancestor of the Pampean WMacrauchenia, 
and also à family which, as AMEGHINO long ago pointed out, has paral- 
leled the horses in the most wonderful way. Certain of these genera are 
tridactyl, with the lateral digits reduced to dew-claws and the weight 
supported entirely upon the median or third digit, giving to the foot a 
most striking and deceptive resemblance to that of Protohippus and 
Hipparion. Another genus even surpasses Æquus in the completeness of 
its monodactylism, the lateral metapodials being reduced to minute pro- 
ximal rudiments, far smaller than the splinthbones of the horse. Here 
again the carpus and tarsus show that these animals are not related to 
the perissodactyls, but to be much more nearly allied to the Toxodontia. 
They afford one of the most remarkable and instructive examples of pa- 
rallel or convergent development known among mammals, but, at the 
same time, they demonstrate that in so complicated a structure as the 
mammalian skeleton convergence or parallelism can never result in 
complete identity, but merely in a greater or less number of striking 
similarities. 
The problem of relationships between these South American ungulates 
and those of the northern hemisphere cannot yet be definitely solved, 
but it is my present opinion, derived from an examination of the pre- 
Santacruzian types of hoofed animals, that all the South American 
groups are more nearly related to one another than to any northern 
group, and that both the northern and southern types go back to a 
common ancestry, Which is nearly represented by the Condylarthra. 
OsBorx's recent discovery of a true armadillo in the Bridger Eocene of 
North America proves the possibility of a communication, however 
roundabout and indirect, between the two Americas in late Mesozoic or 
early Tertiary times, so that the existence of Condylarthra in the sou- 
thern continent is what we should expect. 
