466 PROCEEDINGS OF THE NATIONAL MUSEUM. vol.40. 



of the structural diversities inside this group, perhaps because of the 

 uncertainty that still attaches to the original genus Cryptodesmus 

 and the family Cryptodesmidse. 



Comparison of African material with the original specimens of 

 Cryptodesmus in the Berlin Museum showed that the reference of 

 African species to this genus was without warrant, and the presence 

 of many constant structural differences among the members of a rich 

 African fauna led to the recognition of a considerable series of new 

 genera, and finally to the grouping of these genera into families. It 

 was also found that some of the tropical American millipeds were 

 much more related to the African genera than the first species that 

 Peters assigned to Cryptodesmus, so that some new American genera 

 were established and assigned, provisionally at least, to African 

 families. 



As the American fauna becomes better known it appears more and 

 more probable that the original genus Cryptodesmus stands well apart 

 from the other members of the series of forms usually associated with 

 it. Indeed, in Pocock's recent treatment of the Central American 

 millipeds in the Biologia Centrali-Americana a new family Peridon- 

 todesmidae is established for the genus that seems to be more nearly 

 related to Cryptodesmus than any other member of the Central 

 American fauna. The first segment of Peridontodesmus is not ex- 

 panded as in Cryptodesmus and the carinas are broader and more 

 deeply notched, but such differences would not be considered very 

 serious if other features were alike. 



It may be that the rough outlines of the first and last segments of 

 the type of Cryptodesmus shown on pi. 60, figs. 2« and 2c, will help to 

 explain the tendency to look upon any 20-segmented tropical milliped 

 with a large first segment as a relative of Cryptodesmus, and to show 

 at the same time that such conclusions are generally unwarranted. 

 Apart from the broadly expanded first segment, the original type of 

 Cryptodesmus does not show any tendency toward either of the two 

 forms of specialization that characterize most of its supposed rela- 

 tives. The segments do not have radiating marginal areas as in the 

 Pterodesmidae, nor enlarged dorsal tubercles, crests, or processes as 

 in the Stylodesmidae, Hercodesmidae, and Chytodesmidae. The 

 repugnatorial pores are equally unspecialized. They are not located 

 on the anterior part of the segment as in the Pterodesmidae, nor on a 

 special lobe or tubercle as in the Stylodesmidae and their allies, but 

 are located near the margin on a slight elevation, somewhat as in 

 Scytonotus. The segments have simple piliferous tubercles as in 

 Peridontodesmus and Scytonotus, and in the pore characters also 

 there is more agreement with these genera than with the Pterodes- 

 midae or the Stylo desmidae. There would seem to be better justifi- 

 cation for the recognition of the family Scytonotidae than for the 



