556 THE PANORPOID COMPLEX, iii., 



tion, of course, would be M 3 , JVJ 4 and M 5 respectively for these 

 three branches; and, as this is not an Odonatological research, I 

 propose to use these terms in this paper. 



A careful analysis of the Odonate arculus shows that it is 

 really divisible into three parts, viz.: — 



( i . ) the uppermost part, from the point where M leaves R to 

 the point where M. 1 ^ diverges from M 5 



(ii. ) the middle part, consisting of a short basal piece of M 5 

 from its divergence from M 1 _ 4 to the top of the cross-vein. 



(iii.) the lowest part, formed from the cross-vein, called the 

 posterior arculus. 



The middle portion (ii.) is apt to be overlooked, since, in the 

 majority of recent Odonata, M 1—4 and M 5 come off together 

 from the same point on the 'arculus; or, in other words, the 

 cross-vein forming the lowest portion has moved up to the point 

 of origin of M 5 , thus obliterating (ii.) altogether. But in 

 some fossils this portion is the longest of the three, as, for 

 instance, in Aeschnidiopsis (22) . This middle part is of the 

 greatest importance in the present discussion, and must not for 

 one moment be lost sight of. 



Lying distad from the posterior arculus, in the Zygoptera, is 

 the discoidal cell, or quadrilateral, whose upper and lower sides 

 are portions of M 5 and Cu respectively. The condition that 

 this cell can be formed was laid down by me in a former work 

 (21, p. 57), viz. that the bifurcation or forking of Cu must 

 take place at some point distad from the level of the arculus. 

 If tbe cubital fork were placed at some point closer to the base 

 than the arculus is. no discoidal cell could be formed, and the 

 evolution of the arculus would have to proceed along different 

 lines . 



Now, in all archaic types within the Panorpoid Complex, the 

 cubital fork lies very close to the base, and the arculus lies 

 at a level somewhat distad from it. Hence it was not possible 

 for a discoidal cell of the Odonate type to be formed, and the 

 evolution of the arculus has proceeded upon different lines. 

 Instead of remaining separated by a definite area, out of which 

 in the Odonata, the discoidal cell and the area distad from it 

 are formed, M 5 and Cuj have become more and more approxi- 

 mated, and very early became fused, for at least a considerable 

 portion of their lengths, as we find them already in the Permian 

 Paramecoptera . This condition is shown in Text -fig. 41, b. The 



