216 GEOEGE JOHN EOMANES I88I- 



fertile hybrids. It is by no means enough to point to 

 natural and fertile hybrids as cases opposed to phy- 

 siological selection unless it has been shown by 

 experiment through a generation or two that these 

 hybrids are fully fertile — i.e. as fertile as their parent 

 species. Now, experiments of this kind have rarely 

 been carried through. If you assume that the result 

 of carrying them through would be destructive of 

 physiological selection by proving that fertile hybrids 

 are, as a rule, fully fertile, and also (which is very 

 important) that in any cases where experiment may 

 show them to be so, further experiment would 

 fail to show that isolation has not been effected in 

 any other way (as by pre-potency, differences of 

 insect fertilisation, &c.) — in short, if you assume 

 that fertiHty is as complete between the two asso- 

 ciated species as it is within each species, how is it 

 conceivable that they should continue to be distinct ? 

 In this connection it is well to consult Gulick's paper 

 already referred to (especially p. 259, paragraph 1st) 

 on the theoretical side, and Jordan's papers and 

 books on the practical side. I have repeated the 

 latter's observations on poppies, and find that where 

 any considerable number of individuals are concerned, 

 natural selection is not nearly so great a power in 

 this respect. (Even in cases where it happens that 

 in-breeding is necessarily confined to single herma- 

 phrodite individuals for numberless generations, the 

 handicapping is not fatal : witness flowers which 

 habitually fertilise themselves before opening — es- 

 pecially some species of orchids, which never seem 

 to do otherwise, notwithstanding the elaborate pro- 



