48 THE MEDUSAE. 



canal system, for not only do species otherwise closely allied differ in this 

 respect (e. g. in the genera Cunoctantha and Cunina), but different genera 

 may exhibit gradations in the degree of development of the canals ; so that 

 it is impossible to draw any sharp line between those in which they are 

 developed and those in which they are rudimentary. Furthermore, it has 

 already been shown conclusively that in at least one species with canals, 

 i. e. Cunina proboscidca Metschnikoff, a secondary generation of medusae 

 without canals is developed, — an observation I have been able to repeat on 

 a second species, Pegantha smaragdina, sp. no v. A further criticism is that 

 it is, to say the least, questionable whether either Aegina or Aeginura, 

 included by Vanhoffen in the Diocheteumena, ever possesses a well-developed 

 canal system, Maas (: 05) having found it rudimentary in both. The use of 

 the mode of development as a basis for classification in this order would seem 

 at least premature, since as yet in only three genera (Cunoctantha, Cunina, 

 and Pegantha) do we know positively that it is indirect ; and if this feature 

 should be employed, it is incompatible with Vanhoffen's scheme, since one of 

 the species in which it is known to be indirect, Cunoctantha odonaria McCrady, 

 which should therefore belong to the Diocheteumena, has been shown by 

 the miscroscopic studies of Wilson ('87) to lack any trace of a canal system, 

 a character placing it in the Adiocheteumena. 



A much more important character, since it is more constant and appar- 

 ently more closely correlated with the natural relationships between species, 

 and one on which Maas (: 04% : 04°, : 05) has laid special stress, is the presence 

 or absence of gastric pockets. 



It is true that there is some evidence {e.g. in Cunina proboscidca) that these 

 structures, like the canals, may be present in one generation while absent in 

 tin' other. But this evidence is not altogether conclusive, since it is now well 

 known that in at least one species, Cunoctantha odonaria, the pockets are 

 not formed until a stage in growth much more advanced than that to which 

 either Metschnikoff (\S6 a ) or Stchelkanowzcw (: 06) have traced the secondary 

 generation of Curma proboscidea. Vanhoffen (:07) has himself employed this 

 character to separate two families, Aeginidae and Peganthidae, under the 

 suborder Diocheteumena (with peripheral canal system), but he has entirely 

 neglected il in the second suborder, Adiocheteumena, apparently believing 

 thai Solmaris may have gastric pockets, whereas all previous students agree 

 tli.it in this genus, as well as in the well-known Mediterranean Solmonda 

 flavescena, the conditions of the gastric cavity exactly parallels that in Pegan- 



