190 MR. E. J. TILLYARD ON THE RECTAL BREATHrNG-APPARATUS 



phylefcic line of ascent o£ the Foliate Tyjye, rather the latter, as formed in 

 ^schna and Anax, must be considered to have arisen from a somewhat 

 simpler development of the Implicate Type, 



9. The Foliate Type represents the highest development of the Duplex 

 System within the ^schnince. The papillo-foliate sub-type is a specialization 

 from the normal foliate sub-type of ^FJschna. 



10. The Lamellate Type is an isolated type within the Duplex System, of 

 very high specialization, but undoubtedly also of very great antiquity. Its 

 point of phyletic union witli the other gill-types must be sought for far back 

 in the early days of the Undulate Type, before the main folds attained any 

 great development. In the formation of the Lamellate Type the main folds 

 became completely suppressed, the cross-folds became highly developed as 

 gill-lamellse, and the basal pads underwent a correlated localization as flat 

 supporting discs and became deeply coloured. 



11. The number of undulations and corresponding sets of cross-folds 

 corresponded primarily with the number of primary efferent tracheae de- 

 veloped, and was probably at the first from six to nine. This number early 

 increased to twelve, the number still found in the mature larvse of Petalura,. 

 Corchdeg aster, Austrogomphus, HemicfompJius, Anax, and Synthemis. Further 

 development increased this number up to as much as thirty in highly 

 specialized cases (e. g., Diplacodes). 



12. Within the Lamellate Typ>e only minor variations of structure occur. 

 We can perhaps distinguish as sub-types the Archilamellate gills in the 

 Synthemini and the Neolamellate gills in the rest of the Libellulidse. Within 

 the latter sub-type some minor differences separate the CorduUince from 

 the Libellidina'. The latter subfamily shows the higher specialization,, 

 culminating in Diplacodes. AustrocorduUa has struck out a side-line of 

 specialization by an attempt to develop papillse. Cordulepliya appears, as 

 in other larval characteristics, to belong definitely to the Eucordidiini. 



The above twelve conclusions are illustrated by the phylogenetic diagram 

 in text-fig. 21. 



The value of the evidence afforded by the gill-structure in elucidating the 

 general phylogeny of the Anisoptera is considerable, and tends to strengthen 

 the views already put forward by recent workers in that field. In pointing 

 to Petalura and Cordidegaster as the most archaic of living Anisoptera, it 

 agrees with the conclusions already obtained both from Palaeontology and 

 from other branches of the Comparative Morphology of both larva and 

 embryo. The greater antiquity of AuHrogomp)lms as compared with Hemi- 

 gomplius, and hence also with Gomplius and Onychogomplius, is a very 

 interesting point, though only of secondary importance. Passing on to 

 the jFiSchnince, it is much to be regretted that a larva of the archaic tribe 

 Petaliini cannot be obtained for study, since this will almost certainly either 

 give us the missing link between the Implicate Type and the Undidate Type^ 



