150 Animal Life and Intelligence. 



variations. This view, however, would seem to involve an 

 hypothesis which may be true, but which, in any case, 

 should be indicated. For it is clear that if new or favour- 

 able variations arise in this *way, the germinal union 

 cannot be a mere mixture, but an organic combination. 



An analogy will serve to indicate the distinction implied 

 in these phrases. It is well known that if oxygen and 

 hydrogen be mixed together, at a temperature over 100 C., 

 there will result a gaseous substance with characters inter- 

 mediate between those of the two several gases which are 

 thus commingled. But if they are made to combine, there 

 will result a gas, water- vapour, with quite new properties 

 and characters. In like manner, if, in sexual union, there 

 is a mere mixture, a mere commingling of hereditary 

 characters, it is quite impossible that new characters 

 should result, or any intensification of existing characters 

 be produced beyond the mean of those of ovum and sperm. 

 If, for example, it be true, as breeders believe, that when 

 an organ is strongly developed in both parents it is likely 

 to be even more strongly developed in the offspring, and 

 that weakly parts tend to become still weaker, this cannot 

 be the result of germinal mixture. Let us suppose, for 

 the sake of illustration, that a pair of organisms have each 

 an available store of forty units of growth-force, and that 

 these are distributed among five sets of organs, a to e, as 

 in the first two columns. Then the offspring will show the 

 organs as arranged in the third column.* 



Parents. Offspring. 



10 

 9 



7 



40 40 40 



* Latency is here neglected. Mr. Francis Gallon has shown, statistically, 

 that the offspring, among human folk, inherit from eacli parent, T ' s from 

 each grandparent, and the remaining J from more remote ancestors. In 

 d >mesticated animals, reversion to characters of distant ancestors sometimes 

 occurs. This, however, does not invalidate the argument in the text, which 



