28 
caudal vertebrae are alone developed, and ultimately give 
rise to tlie " hypural " plates which constitute the bony 
framework of the tail. The upper or neural arches are nearly 
suppressed, and thus the chorda dorsalis is directed towards 
the upper angle of the tail fin, where it may be detected in 
the adult fish, though partly concealed by a short cartilaginous 
sheath. 
The fishes ordinarily reckoned as homocercal are therefore 
excessively heterocercal, the upper lobe of the tail being sup- 
pressed. This recent discovery need not in itself destroy the 
classificatory value of Agassiz' distinction, could it be still 
maintained that the external outline of the tail is a safe • 
guide to affinity. In a shark the upper lobe is predominant, 
in a salmon the lower. Though both are heterocercal or 
unsymmetrically developed, the resulting contours are quite 
unlike. Had there ever been much substantial truth in the 
distinction of homocercal and heterocercal as a clue to the 
discovery of real affinity, a change of phrase would have set 
all right. But tail characters are not available for the iden- 
tification of any of the great groups of fishes. At most 
they are of generic value. Within the order Ganoidei we 
have a great variety of structure in this respect. Polypterus 
and many fossil genera are " diphycercal,'* the nearest 
approach to true symmetry found among fishes ; Lepidosteus 
and the sturgeons, besides fiaany fossil forms, are visibly 
heterocercal. Calamoichthtjs has an urostyle developed on 
one side only of the caudal fin. No appeal to analogy would 
completely decide whether such a fish, for example as 
Holoptychiiis, whose caudal fin is not accurately known, had 
a symmetrical or unsymmetrical tail, nor would analogy 
suffice to show how the vertebral column terminated in such 
a fish. 
Just as a classification based upon tail characters would 
break up the well-defined order of Ganoid Fishes, so also the 
