NARCOMEDUSAE. 49 



tha, there being no true gastric pockets at any stage in development. 

 Finally, I may call attention to the fact that the employment by Vanhoffen 

 of a single character, i. e. number of tentacles as the sole basis of generic 

 separation in this complex group, has resulted, as might have been ex- 

 pected, in such unnatural combinations as Cunoctantha with Aeginura and 

 Cunina with Aegineta, genera with which they are but distantly related as 

 is shown by the condition of the gastric pockets, and by the otoporpae; and 

 of Solmissus with Solmaris, genera sharply separated by the presence of 

 gastric pockets in the former and their absence in the latter. 



On the whole, the scheme proposed by Maas (:04 a , :04 c ) much more nearly 

 represents what I believe to be the probable natural relationships among 

 the Narcomedusae. According to Maas Narcomedusae in which gastric 

 pockets are present are to be subdivided into two families, Cunanthidae and 

 Aeginidae, according as the pockets are radial and equal in number to the 

 tentacles, or adradial and twice as numerous as the latter; but as to whether 

 the latter condition is actually derived from the former by the secondary divi- 

 sion of pockets primarily radial, it is perhaps best to reserve judgment, since 

 the present collection contains one species with interradial pockets equal in 

 number to the tentacles, a condition hardly to be derived from one with 

 radial pockets. Opposed to these two families are the Solmaridae, in which 

 there are no gastric pockets. 



There are, as Maas (: 04) points out, strong reasons for believing that the 

 two families with gastric pockets are more nearly related to each other than 

 is either of them to the Solmaridae ; and Dr. A. G. Mayer, who holds this 

 same view, has suggested to me the desirability of uniting all Narcomedusae 

 with gastric pockets into a single family, reducing the Cunanthidae and 

 Aeginidae to the rank of subfamilies. I doubt, however, whether to take 

 this step would result in advancing our knowledge of the true phylogenetic 

 relationsbips of the various members of this complex group; for although 

 the similarity of the Cunanthidae as a whole to the Aeginidae is no doubt 

 close, yet Solmissus, a cunanthid, so closely simulates certain Solmaridae, 

 e.g. Solmaris and Solmoneta, in the peculiar form of its sense organs, that it 

 is hard to believe that the resemblance is accidental. In all probability our 

 knowledge of the development of the Narcomedusae will be greatly widened 

 within the next few years, with the result- of giving a much firmer basis for 

 revision than is possible from the descriptions of the adull stages as yet 

 published. In the meantime it seems tome that we may well retain Maas's 

 provisional diagnosis of the families composing the order, as follows: — 



