Carpenter — Relationships between Classes of ArtJiropoda. 329 



closely with the similar structures in the Collembola. And by- 

 dissection of the head, I have succeeded in demonstrating the presence 

 of a pair of minute maxillulfe^ associated with the tongue, and lying 

 between mandibles and maxillae, just as they do in the Springtails 

 (fig. 3, mxL). All that is now wanting to bring the segmentation of 

 Scolopendrella into perfect agreement with that of the primitive Insects, 

 is embryological proof of the presence of the vestigial tritocerebral 

 appendages of the head. With confidence, therefore, we may postulate 

 a Scolopendrelloid ancestry for Insects, Centipedes, and Millipedes. 



Among the Centipedes we find very great variation in the number 

 of the body-segments, Lithobius and Scutigera having only fifteen pairs 

 of walking-legs, Scolopendra and its allies twenty-one or twenty-three 

 pairs, and the Geophiloids often more than a hundred pairs. Now, 

 according to the view of Haase, adopted by BoUman ('93), the fifteen- 

 legged groups must be regarded as the more primitive on account of 

 the comparative simplicity of the tracheal system in Lithobius, the 

 spiracles having no closing apparatus ; and it is especially noteworthy 

 that a correspondingly simple stage is passed through in the develop- 

 ment of the Scolopendi'id^. Although the tracheal system of Scutigera 

 is highly specialised and the spiracles dorsal in position, the head and 

 mouth-parts of that animal retain many primitive characters. 



But the important discovery recently announced by Pocock ('02) 

 of a Tasmanian genus of Centipedes (Craterostigmus) with fifteen 

 pairs of legs like a Lithobioid, and twenty-one tergites like a Scolo- 

 pendra, is believed by him to indicate the descent of the Lithobioids 

 from Scolopendroid ancestors through the loss of sis segments — the 

 3rd, 6th, 9th, 11th, 14th, and 17th. He suggests that, in Cratero- 

 stigmus, the tergites of these segments are still retained. Two objections 

 may be made to this view. It is hard to imagine a reduction in the 

 number of segments by the loss of a scattered series such as this. 

 And the derivation of the Lithobioids from the Scolopendroids, through 

 Craterostigmus, would destroy the remarkably close correspondence 

 between the position of the spiracles on the 1st, 3rd, 5th, 8th, 10th, 

 12th, and 14th body-segments of Lithobius and the corresponding 

 segments (except the 1st) of Scolopendra. As Craterostigmus exhibits 

 several Geophiloid characters, it is more likely that its six "minor" 

 tergites should be compared to the smaller sections of the incom- 

 pletely-divided dorsal plates of the GeophilidjB. 



' While this paper is passing through the press, I find that the maxillulse of 

 Scolopendrella have heen seen and figured hy Hansen ('03). 



