THE SPEED OF RETROGRESSION 203 



harmful to the race. Selection may be compared to a sculptor 

 who designs and rough hews a statue ; sex to a colleague, more 

 skilful mechanically, who chisels out the finer lines. 



334. We never see racial progression except when selection, 

 natural or artificial, is at work. The evidence is, that given 

 sufficient time, retrogression invariably occurs when there is 

 cessation of selection, its speed being proportionate to the recency 

 of the antecedent progression. Now, when a character (e.g. limb) 

 has been long established, constant interbreeding has so blended 

 together all the strains of a race, that mating individuals differ 

 comparatively little. At any rate, apart from mutations, which 

 are comparatively rare and tend to be eliminated, individuals, 

 exhibiting every gradation between the extremes of ' normal ' 

 variations, occur the most numerous type and therefore the most 

 influential as regards inheritance being those which approximate 

 most closely to the specific mean. In such cases, therefore, the 

 retrogression which follows the cessation of selection depends 

 wholly on the prepotency of new retrogressive variations, and is 

 a very gradual process. But when the character (e.g. any of the 

 special points of prize breeds) has been newly established, mating 

 individuals differ more, for the blending will have been less 

 thorough, and retrogression, therefore, is more rapid. It is possible, 

 indeed I think it is probable, that retrogression follows mere 

 cessation of selection even in parthenogenetic species. But the 

 prolonged persistence of useless structures in such rapidly repro- 

 ducing types as Cypris reptans, as well as the fact that partheno- 

 genetic species tend to break into a multitude of progressive 

 ' varieties,' indicates that retrogression on cessation of selection 

 is a slower process than when blending is present. 



335. The higher animals are the most complex products of 

 evolution. The multitude and differentiation of their structures is 

 very great, and the fact that they are active moving machines renders 

 necessary a very delicate co-adaptation of all their parts. A tree 

 may differ, always does differ, greatly in shape from other members 

 of the species ; but, if a higher animal diverges much from his type 

 in shape, he is a monster and cannot persist. But this very com- 

 plexity of structure and closeness of adjustment increases the area 

 of variability and the chance that variations will throw the whole 

 machine out of gear. It is significant that the reproduction of all 

 these animals is bi-parental. The blending of useful characters, 

 combined with the automatic planing away of redundancies, relieves 

 Natural Selection of half its task of adapting them to the environ- 



