SILICIOUS AND HORNY SPONGES WILSON. 357 



only a particular combination. Thus, on this idea, it would be logical 

 to divide the species now grouped under TetiUa into three genera. 



But another difficulty faces us here in that the extremes of a char- 

 acter, the sharply contrasting conditions that constitute the members 

 (allelomorphs) of a pair, are so often connected by intergrades. 

 Thus in TetiUa between nonfibrous cortex and fibrous cortex there 

 are intergrades. And it may confidently be said that the more in- 

 tensive grows the study of a group of species, the more of these in- 

 tergrading series come to be known. Hence it may be expected that 

 the characters above designated as b and B will, like a and ^4, be 

 found to intergrade, and c and C likewise, if indeed the recorded 

 data do not already justify that conclusion. 



In view of the common existence of intergrades between the con- 

 trasting members of a pair of generic characters (a and A, for ex- 

 ample) , it becomes in the end impossible to split sponges, organisms 

 in general probably, into genera, each of which shall represent a par- 

 ticular combination of characters. 



What practice remains then for the classifier who knows that 

 without systematics biological data in general can not be recorded 

 with any certainty that they will be found again, and who therefore 

 must classify, but who does not wish to set up a system of categories 

 which can be precisely defined only because they are artificial and 

 into which, while some organisms go nicely, others can be brought 

 only after a character is pared down in thought, or extended in 

 thought? One way out, and, as I have indicated above, the way 

 into which we have fallen in the case of TetiUa, is to recognize larger 

 heterogeneous genera ( TetiUa, for example) and other smaller, more 

 homogeneous ones. I can not see at present any better way. It is 

 certainly a preferable method — that is, one that enables biologists 

 to find the recorded data more successfully — than that of building 

 up genera, all of which overlap extensively. 



Similar considerations apply to subgenera. Where a genus, TetiUa 

 for instance, becomes very large and heterogeneous, some grouping 

 of species becomes necessary. And yet because of the intergrading 

 of characters a sharp division of the whole into subgeneric groups 

 is probably only temporarily possible. Forms will, for instance, 

 doubtless be found with pore areas of such a kind as to constitute 

 these forms intermediates between the subgenus proposed below, 

 Cinachyrella, and species the whole dermal membrane of which is a 

 sieve membrane (as in TetiUa ciliata, for example). I would pro- 

 pose to leave such intermediates in the body of the genus, which 

 thereby still remains heterogeneous. 



Subgenera used in this sense are not really groups into which (all) 

 the species of a genus are divided, but rather groups which are 

 sufficiently homogeneous to be set off from a still heterogeneous 

 remainder. The advantage in this treatment, if there is one, and it 



