RETROGRESSION 533 



the type would rise : there seems no escape from the continual rising at 

 any rate so far as the critical limit D V. But beyond this limit, natural 

 selection ceases to work, and ancestral influences act unchecked. Hence 

 LQ will tend to move nearer OP, i.e. in the direction LR. This is the 

 second cause for the depression of the path noticed above. 



55. Retrogressive tendency. But as yet we have found nothing to 

 arrest continual and indefinite departure from the original type up to 

 the point V. There is nothing which can permanently balance natural 

 selection, so that the type may remain stable in the presence of its natural 

 enemies. And yet our experience is that such stability is reached. If 

 we admit the existence of such stability, then we may put the inference 

 as follows : Let any generation consist of A individuals who are destroyed 

 before breeding, and B individuals who survive and have offspring. If a 

 condition of stability is reached the next generation will be the same 

 A unfit and B fit. But all these are by hypotheses the offspring of the 

 B members only of the preceding generation. These B members do not, 

 therefore, produce offspring B, fit like themselves (on the average), but 

 partly B and partly A ; the average of the offspring is lower than that of 

 the parent : and this goes on continually. The parents are of a constant 

 type B : the children are of a constant inferior type between B and A. 

 If the difference remains really constant, ancestral resemblance cannot 

 explain it, for this is all the time growing weaker. We seem to be thrown 

 back on some innate tendency to retrogression, which may possibly be a 

 universal tendency. We assumed at the outset that children might deviate 

 from their parents in either direction : but the deviations may not be 

 equal. We should have a vera causa of the kind we are looking for if the 

 tendency to retrogression were the stronger. It is not necessary that it 

 should be always stronger; it would suffice if it became stronger with growth. 



56. Dr Archdall Reid has, however, no hesitation in stating the 

 principle in a universal form. He wrote to me on July 16, 1909: "In 

 all structures retrogressive variations tend to predominate, and do in the 

 long run predominate over progressive variations. Therefore we get 



(a) Retrogression on relaxation of selection, and more especially on 

 cessation of selection. 



(b) Stability of type when selection balances the tendency to 

 retrogression. 



(c) Progression when the strin- 

 gency of selection is sufficient to 

 overcome the tendency to retro- 

 gression (reversion)." 



He adds that 



" Retrogressive variations tend 

 to be prepotent in a blend, i.e. the 

 retrogressive parent is on the average 

 better represented in the blend." 



57. I am not clear whether this is really an additional effect. 

 Suppose OP X and OP 2 to represent the parents, OP 2 being the retro- 

 gressive. Then according to the law of individual retrogression the child 



