TRIASSIC FISHES FROM SPITZBERGEN 
H7 
over the ethmoidal region according to the pattern still seen more or less distinctly 
persistent in Dictyonosteus, Eusthenopteron and Coelacanthids. Moreover, the bone-plates 
of the cheek must also have consisted of a larger number of independent elements 
than is found in, e. g., Holopthychius (cf. Watson and Day, 1916, pp. 2—7, 22, fig. 1).— 
In other words, in this theoretical ancestral form of Tetrapods and Crossopterygians 
the dermal skeleton in the cranial roof and cheek would have consisted of a rather 
large number of fairly small plates, arranged in a certain regular way. The majority 
or perhaps all of the elements for which we now use the same names in Tetrapods 
ahd fossil Crossopterygii would not be inherited as such from a common ancestral 
form, but might have arisen independently in both cases. 
If my view with regard to the common ancestral form of the Rhipidistids and Tetra¬ 
pods is correct, it also implies that the Tetrapods were developed very early probably 
in lower Devonian or even in upper Silurian (cf. Sromer, 1920; Broili, 1918, p. 165). 
Concluding remarks on the systematic position of the Coelacanthids. 
I have been able to show in this work the untenability of the view maintained 
by Goodrich 1909 and to some extent by Watson (1912) as to the mutual relations of 
the Coelacanthids and the Rhipidistids. On the other hand Goodrich’s assumption as 
to the relationship of the Polypterids has partly received additional support, and because 
of this I have considered that I ought to separate these from the fossil Crossopterygii 
and place them as an independent order between the latter and the Actinopterygii. My 
suggestion for a provisional classification will therefore appear as follows: 
Order Crossopterygii 
Sub-order Haplistia 
Fam. Tarassiidae 
Sub-order Rhipidistia J ) 
Fam. Holopthychidae 
» Rhizodontidae 
» Osteolepidae 
» Onychodontidae 
Sub-order Actinistia 
Fam. Coelacanthidae 
Order Brachiopterygii 
Sub-order Cladistia 
Fam. Polypteridae 
Order Actinopterygii. 
*) Watson and Day. (1.916, p. 26) point out the great difficulty now present with regard to the subdivision of 
the Rhipidistids into families. The prevalent classification in this request has not turned out to correspond to the 
natural relationship between the different forms. Both these investigators consider it most convenient at present to 
refrain provisionally from any attempt at a classification into families. Birks (1916, p. 327) also came to the con¬ 
clusion that the subdivision of the Rhipidistids hitherto used is unsatisfactory. 
19* 
