RECAPITULATION 59 



only another name for reversion is not clearly recognized in 

 biological literature. But if development is by recapitulation 

 as it certainly is it is difficult to imagine what else it can 

 be. It is in fact impossible to conceive of a case of regression 

 which is not one of reversion, or vice versa. 1 No doubt it is 

 often not easy to recognize regression as reversion, but that 

 is only because our vision is obscured by the complex con- 

 ditions under which the problem is usually presented to us. 

 When reversion occurs on a great scale, when the whole 

 organism or a considerable part of it reverts to the ancestral 

 type, and what is known as atavism occurs, we recognize the 

 reversion but not the regression. When the reversion affects 

 only part of a structure, and is disguised by progressive 



1 The fact that regression is identical with reversion is the main 

 contention of the present and the three succeeding chapters. In the 

 discussion of details the central truth is apt to be lost to sight. It is 

 needful, therefore, even at the cost of tedious repetition, to give it all 

 possible emphasis at the outset. It would be well if the reader now 

 paused awhile and tried to imagine an instance of regression which is 

 not also an instance of reversion. He will find it impossible. Every 

 instance of regression is a failure to recapitulate part of the parent's 

 development, and, consequently, part of the life-history. It is there- 

 fore, of logical necessity, an instance of reversion. We may illustrate 

 the argument by taking an extreme example. The first digit on a limb 

 has been evolved by the superimposition of progressive variation on 

 progressive variation in an immensely long series during an immense 

 number of generations. Suppose, now, a child of normal parents is 

 born lacking the first digit. Then his variation if not an example of 

 latency (see 101 106), which properly speaking is not regression 

 is a regressive variation. It is also an act of reversion. As regards this 

 one character he has lapsed the whole of the progressive variations made 

 by millions of his ancestors. He has reverted in this one particular to 

 that remote ancestral condition which existed previously to the evolu- 

 tion of the digit (or even of the limb). It matters not that very prob- 

 ably no ancestor ever existed who had four perfectly developed digits, 

 and only four. It is not contended that the child has reverted as regards 

 the whole hand ; it is only contended that he has reverted as regards 

 the first digit. Suppose again instead of completely lacking the first 

 digit the child were born with a malformed rudimentary digit, a mere 

 stump. Here, again, we should have reversion and regression, though 

 in a form less complete. The digit may not, and probably would not, 

 closely resemble anything which existed in the ancestry ; but that 

 would only mean that different parts of the digit had reverted unequally. 

 A digit like an entire limb is made up of parts, and it does not neces- 

 sarily or usually revert in all its parts equally. The reason why 

 regression so seldom presents the appearance of reversion will be more 

 fully discussed presently (see 132 and 148) ; but meanwhile, since 

 development is by recapitulation, there is no escape from the conclusion 

 that every act of regression must be an act of reversion. It should be 

 noted, however, that when a child lacks a digit this variation is much 

 more probably due to latency than to reversion. Reversion to an 

 ancestor so extremely remote is usually a slow process. (See 139.) 



