BV R. J. TILLVARD. 723 



branched off from the Corduliince much further l)ack than we arc 

 iuchned to suspect, and that, in the resemblance between Group i. 

 of the Libellulinoi and such forms as Cordulephya, we have only 

 a imrtial solution of the problem. These two are nearer by asthe- 

 nogeneiic convergence at the present day than their ancestors pro- 

 bably were before as theno genesis began to act upon them. Never- 

 theless, those ancestors did possess archaic characters common to 

 the original stock, and it was their failure to improve on the weak 

 points of their venation al structure, that caused them to drop 

 behind in the race of progress. The main body of these ancestral 

 forms has long ago perished, but there remain to-day those few 

 "end-twigs" of the old stem, which have survived by adopting 

 asthenogenesis. This line of development must, of necessity, pro- 

 duce forms that are true convergences, in the sense that they now 

 appear more closely related than they would, if we knew their past 

 history in full. 



To deal, finally, with the problem of the Libelluline anal-loop, 

 we have to state one very strong objection to the theory of its re- 

 development from a narrowed-down type of hind-wing. It is this : 

 those who support this theory will scarcely admit the possibility of 

 the re-formation of the unstressed four-sided "triangle" from a 

 more highly-developed three-sided structure. They say, that, when 

 once so useful a form as the latter had become developed, no 

 species could possibly have gone back to the ancestral form, or it 

 must have perished. But they ask us to believe a far harder thing ; 

 viz., that a race with a broad hind-wing, whose anal area was built 

 up roughly on lines similar to Petalura, could not only lose this 

 structure (whose uselessness they surely would not attempt to 

 prove, in the face of the survival of the Aeschnine and Macromian 

 loops), 'but also that, having lost it, they redeveloped a new type 

 of loop afresh. Surely this theory is only supported because of 

 the apparent necessity of insisting on the archaic value of the 

 narrow-winged types. This Libelluline loop (like all other loops, but 

 not necessarily on the same plan) developed from the original 

 fairly broad hindwing of the Anisopterid stock, which contained 

 all the material (in the form of numerous small cells) for the for- 



