VARIATIONS RESEMBLING REVERSIONS 129 



mean by reversion the re-expression of an ancestral character after 

 a period of latency, it is obviously a particular mode of inheritance. 

 From another point of view it is a variation, and due to some 

 unknown germinal conditions which permit a long-latent, but 

 never lost, character to re-assert itself. When we consider the 

 intricate reductions which occur in the maturation of the germ- 

 cells, and the not less intricate reinforcements involved in 

 amphimixis, it is not impossible to imagine how an ancient 

 latent character may come to the front again after many genera- 

 tions. 



But we have also to remember that, apart from reasser- 

 tions of what is relatively old, there is a continual emergence 

 of what is relatively new. What occurred once as a new variation 

 may occur again, and it is a certain fact that the same type of 

 variation occurs over and over again in varieties of different 

 species. How many red and blue flowers have white varieties ! 

 how many trees have weeping varieties ! how many Arthropods 

 show similar increase or decrease in the number of their joints! 

 how many birds show albinism ! There are limits to the varia- 

 tions of the kaleidoscope, and to the kaleidoscope of variations. 

 Therefore it is always possible that a variation really occurring 

 de novo, and apart from latent characters, may happen to coincide 

 with an ancestral trait. It may be described as a reversion, but 

 it is really an independent variation. 



Supernumerary mamma5 occasionally occur in human beings 

 in both sexes. Ammon found them in 3 per cent, of German 

 recruits. They obviously suggest the several pairs of mammae which 

 occur in many mammals — e.g. in the half -monkeys or Lemurs. 

 Weismann (1893, p. ^^3) says, " They are undoubtedly to be looked 

 upon as reversions to extremely remote characters possessed by our 

 lower mammalian forefathers." But it seems simpler to regard l^ $fi^ 

 them as independent variations, comparable to many other ab- ^ -L'- 

 normal multiplications of parts. They happen to suggest bygone \^r 

 conditions, but that is probably all that we are warranted in saying. 



Polydactylism in man has been interpreted as a reversion to an 



9 



