THE PURE LINE HYPOTHESIS. 559 



Thus the mean colours of the groups of parents and of the 

 corresponding- red offsprinq- are : 



Mean of A grroup oif Parents 68, and mean of A Offspring: 63 



Mean of B g^roup of Parents 53, and mean of B Offspring- 60 



Mean of C group of Parents 35. and mean of C Offspring 51 



Mean of D group of Parents 17. and mean of D offspring 44 



According to the pure-Hne hypothesis, the means of the 

 dift'erent groups of offspring should remain about the same, 

 instead of showing a distinct fall, corresponding to the drop in 

 the means of the groups of parents ; and it would not be possible 

 to produce various strains O'f colour intensity by selecting from 

 such self-fertilised generations. The utmost effect of selection 

 would be the sorting out of two strains corresponding to the 

 factors for redness in the male and the female grandparent. 

 These factors of the grandparents were closely similar to each 

 other, since the red offspring of each, resulting" from selfing. were 

 not distinguishable in the matter of colour. 



The drop in the means of the parents results from the 

 fluctuating variations in the expression of a character, which 

 according to the views of the majority of Mendelians are non- 

 inheritable. 



It will be noticed that the drop in the means of the offspring 

 is not as great as that in the means of the parents, and such is 

 found almost invariably to be the case, namely, the offspring of 

 abnormal parents tend on the average to be less abnormal than the 

 parents themselves. This falling back to the general mean of 

 the race on the part of the off'spring may be called " regression." 

 but the term must not be confused with " reversion," in which a 

 long-lost ancestral condition suddenly reappears. 



The above method of dealing with the results is elementary 

 and mathematically incomplete ; but it has the advantage of being 

 more readily understood. 



Using Professor Karl Pearson's methods, the following 

 constants were calculated: Mean of parents, 40.4149 units; mean 

 of red offspring, 51.3653 units; standard deviation of parents, 

 20.1130; standard deviation of red offspring, 22.5788; co-efficient 

 of correlation, .435. The correlation table from which these 

 constants v.^ere calculated is given as a record. 



