152 von Ihering, The Dendrocolaptidce. LApril 



Conurinse are distributed in three different subfamilies, the Arainae, 

 Pyrrhurinse and Platycercinse ! 



The same process of development of a certain organ is repeated 

 many times independently in different subfamilies and genera and 

 therefore can be applied only to a limited extent in classification. 



No single organ is of such importance that we can attribute to it 

 absolute preference and it is never possible to determine a priori 

 whether this or that character will be of most importance in syste- 

 matic work. It happens sometimes that a relatively insignificant 

 character will prove of great value, as for example the loss of a 

 remex, which serves as a distinction between the large groups of 

 quincubital and aquincubital birds. The quincubital condition is 

 the archaic one and the loss of the fifth remex although represent- 

 ing a higher phylogenetic degree, must be considered as a process 

 of degeneration, for which it would be stupid to make natural 

 selection responsible. 



What we learn from ornithological studies is that the wide range 

 of variation which leads, or can lead to the origin of new groups, is 

 on the definite lines of evolution which influence also the less im- 

 portant characters but which do not raise any question of survival 

 since both the primitive and modified types succeed equally well 

 in the struggle for existence. 



In more than forty years of uninterrupted biological research I 

 have been unable to discover any facts among free living animals 

 which tend to prove the existence of natural selection, or even to 

 elevate it to the rank of an indispensable or necessary factor in the 

 origin of species. So long as we do not have at our disposal a 

 complete series of morphological and paleontological observations, 

 which would furnish a systematic arrangement of genera on the 

 ground of actual phylogenetic experience, our classifications are 

 more or less a question of our ability to accurately judge the im- 

 portance of morphological characters for systematic use. Barriers 

 erected by anatomists, however celebrated, during the past three 

 decades should no longer be allowed to present difficulties in our 

 ornithological work. 



From the preceding discussion I reach the following conclusions. 



1. The assumed difference between schizorhinal and holorhinal 

 skulls does not exist in the Dendrocolaptidse. The species in which 



