BY R. J. TILLYARD. 177 



We cannot omit, here, a reference to the behaviour of the 

 hindwing-triangle. In Stage 2, that triangle was still distal 

 from the arciiliis. Had it remained so, while Ac receded basad, 

 the anal loop must inevitably have widened, and Synthemis reyina 

 might even to-day then stand as the apex of Anisoj^terous wing- 

 development. But, as soon as the tendenc}' to transverse stretch- 

 ing began, the hindwing-triangle followed the anal-crossing basad, 

 and began to recede towards the arculus, until its basal side 

 finally came to lie exactly under it. It is, indeed, curious to see 

 how the necessity of hroade^iing the hind wing should cause the 

 triangle to recede and tend to broaden out lonyitudinally, while 

 the almost equally insistent necessity for narroivimj the forewing 

 (to prevent overlapping on to the hindwing) should bring about 

 an exactly opposite result, viz., the procession of the triangle 

 away from the arculus, and an excessive broadening of it trans- 

 versely to the wing-axis. In the case of the hindwing, the j9?<// 

 on the triangle stretched it longitudinally; while in the case of 

 the forewing, the push on it, due to the rapid narrowing of the 

 anal area, shut it up transversely like a collapsed framework, at 

 the same time that it drove it away from the wing-base. 



We are now in a position to understand Figs. 9-10. The shape 

 of the area to be supplied by the anal trachea and its branches, 

 has now altered so considerably, that we need not be surprised 

 at the change in the relative importance of the various branches. 

 As this area becomes elongated in a slantwise direction, A2 

 becomes directly attached to Ac, and in this very favourable 

 position, it begins to show a distinct gain in size and length 

 compared with the other branches. Forming, as it now does, a 

 kind of mid-channel along the whole anal area, it supplies the 

 latter with the greater part of its oxygen. On the other hand, 

 Ai, originally more important than Ao, now shows a further 

 slight reduction from the form reached in Stage 2 (Synthemis). 

 It still reaches Cua just below its bifurcation from Cu„ but is 

 an exceedingly slender trachea, very small in comparison with 

 Cuo. It gives off slender branches inwards to help to form the 

 separate cells of the loop, as far as the third set of cells from its 

 base. But the rest of tlie loop, consisting of two more sets of 



