APPENDIX B 367 



moreover, that the actual effect -of bi-parental reproduction 

 (amphimixis) is to cause regression towards the specific (i. e. 

 ancestral) mean; for which reason species that reproduce bi- 

 sexually, and are endowed with large powers of locomotion, pro- 

 duce comparatively few varieties. Such species when they evolve, 

 when they progress or regress, do so as a whole. Unlike par- 

 thenogenetic or self-fertilized types, they form varieties only 

 when sections of the species are geographically separated. It 

 would appear, therefore, that the actual effect, the real function, 

 of bi-parental reproduction is the elimination of useless progres- 

 sive variations. 1 This is necessarily of the greatest importance to 

 large, complex, and mobile animals. For obvious reasons, varia- 

 tions tend to occur in the greatest number in them, and when 

 unfavourable tend to throw their many closely co-ordinated parts 

 out of gear. None of the higher animals, therefore, reproduce 

 parthenogenetically, nor are self-fertilized. Neither the one nor 

 the other method of reproduction provides the maximum quantity 

 of regression. Only some plants and lower animals are thus 

 produced. 



Blended inheritance causes regression to the parental, and 

 therefore towards the specific mean. While it is a potent cause 

 of the elimination of useless progressive variations, it does not 

 prevent that raising of the specific mean as regards useful 

 characters which results from stringent selection. 2 Equally 

 effectual as a cause of regression is that form of exclusive inherit- 

 ance in which one parent is prepotent in the Darwinian sense; 

 for, since this parent is usually the one which has not varied 

 progressively, the ancestral type is reproduced. This form of 

 inheritance is more common and of greater importance than most 

 authors seem to realize. Very seldom, for example, does a child 

 inherit a small variation e. g. a mole which is present in one 

 parent but absent in the other. 



Alternative inheritance cannot cause regression towards the 

 specific mean. The most it can do is to produce new combina- 

 tions of the parental characters new combinations which would 

 have a disastrous effect on the co-ordination of the parts were 

 Mendelian characters less rare. Nevertheless this form of repro- 

 duction does occur; and it should be possible to find a valid 

 explanation of it. 



Alternative inheritance occurs especially in the case of charac- 

 ters in which parents differ greatly, we may even say radically. 

 Thus if one parent be fair and the other dark, the inheritance is 

 usually blended ; or if one parent has a small variation (e. g. a 

 mole), which the other lacks, the inheritance is exclusive in the 

 Darwinian sense ; but if one parent is totally devoid of pigment 



1 See Chapter VI. 



2 See 140. 



