INHERITANCE OF FUNCTIONALLY-WROUGHT CHANGES. 095 



none of them can have more than probability ; while some of 

 them remain, and are likely to remain, very questionable. Ob 

 serve the difficulties. 



(1) The general argument proceeds upon the analogy between 

 natural selection and artificial selection. Yet all know that the 

 first cannot do what the last does. Natural selection can do 

 nothing more than preserve those of which the aggregate charac 

 ters are most favourable to life. It cannot pick out those 

 possessed of one particular favourable character, unless this is of 

 extreme importance. 



(2) In many cases a structure is of no service until it has 

 reached a certain development ; and it remains to account for 

 that increase of it by natural selection which must be supposed 

 to take place before it reaches the stage of usefulness. 



(3) Advantageous variations, not preserved in nature as they 

 are by the breeder, are liable to be swamped by crossing or to 

 disappear by atavism. 



Now whatever replies are made, their component propositions 

 cannot be necessary truths. So that the conclusion in each case, 

 however reasonable, cannot claim certainty : the fabric can have 

 no stability like that of its foundation. 



When to uncertainties in the arguments supporting the hypo 

 thesis we add its inability to explain facts of cardinal significance, 

 as proved above, there is I think ground for asserting that 

 natural selection is less clearly shown to be a factor in the origin 

 ation of species than is the inheritance of functionally-wrought 

 changes. 



If, finally, it is said that the mode in which functionally- 

 wrought changes, especially in small parts, so affect the reproduc 

 tive elements as to repeat themselves in offspring, cannot be 

 imagined if it be held inconceivable that those minute changes 

 in the organs of vision which cause myopia can be transmitted 

 through the appropriately-modified sperm-cells or germ-cells ; 

 then the reply is that the opposed hypothesis presents a corre 

 sponding inconceivability. Grant that the habit of a pointer was 

 produced by selection of those in which an appropriate variation 

 in the nervous system had occurred ; it is impossible to imagine 

 how a slightly-different arrangement of a few nerve-cells and 

 fibres could be conveyed by a spermatozoon. So too it is im 

 possible to imagine how in a spermatozoon there can be conveyed 

 the 480,000 independent variables required for the construction 

 of a single peacock s feather, each having a proclivity towards its 

 proper place. Clearly the ultimate process by which inheritance 

 is effected in either case passes comprehension ; and in this respect 

 neither hypothesis has an advantage over the other. 

 45 



