Oct., i893. THE PROBLEM OF VARIATION. 283 



we shall make much advance towards an explanation of the variations 

 which make evolution. The differences between a large number 

 cf individuals, without reference even to the ontogeny of these 

 differences, are recorded with greater accuracy and completeness 

 than before, but that is all. The same differences might have been 

 present in every generation of the species if it had been created in 

 the form in which we see it. 



In order that evolution may take place, individuals must be 

 generated which are different to any that ever existed before. If we 

 take any particular organ or character, one or more individuals in a 

 generation must have possessed this organ or character in a more 

 advanced (or less advanced in the case of retrogression) condition 

 than any individual which had previously existed. This is the kind 

 of variation which we must account for, and find the cause of, in 

 order to have a complete theory of evolution or of heredity. This 

 must occur over and over again to produce the evolution which, 

 according to the evidence, has taken place. 



Now the actual appearance of new characters has been observed, 

 for instance, in the case of the Ancon sheep, and in wild plants when 

 cultivated ; and, on the other hand, Weismann has proposed the 

 theory that variations of the kind defined above are due to the union 

 of two individuals in sexual reproduction ; but I have not space to 

 discuss, on this occasion, either of these two subjects, because, with 

 regard to the second, Weismann has abandoned the theory, and, with 

 regard to the first, such variations occur at the beginning of the 

 development of the individual. Nothing is more certain than that no 

 theory of variation is worthy of attention unless it takes into account 

 the phenomenon of recapitulation. In a very large number of cases 

 an organ, or a character, or an individual, passes through in develop- 

 ment stages which reproduce, more or less exactly, an ancestral con- 

 dition. In other words, new characters are added usually at the end 

 of ontogeny. We have convincing evidence, then, that the modified 

 individual first resembled its parent, and afterwards became different. 

 The ancestor of the fiat-fish was symmetrical ; but, at some time or 

 other, individuals of this pedigree, after developing into symmetrical 

 fish, became more or less asymmetrical, and every fiat-fish at the 

 present day develops first into a symmetrical fish, and then turns into a 

 Pleuronectid. 



It is certain that no combination of invariable parental germ- 

 plasms could result in this change. How, then, does Weismann's new 

 theory account for it ? The determinants, the units of the germ-plasm 

 corresponding to particular cells, or group of cells, in the adult, are 

 modifiable ultimately by external influences. On this new theory 

 Weismann attempts the explanation of recapitulation in the following 

 words : — 



" The determinants of the id of germ-plasm become endowed with 

 a greater power of multiplication, so that each one of them causes the 



