Correlation and Application of Statistics to Problems of Heredity 95 



from a more complete view of multiple regression, and the more accurate 

 values we now have for the heredity constants. 



The second paper to be referred to was published in The Gardeners' 

 Chronicle for May 15, 1897, and is entitled "Retrograde Selection." In 

 this Galton asks from horticulturists advice as to the cultivation of a 

 plant or plants existing in an original stock R and a stable variety V. 

 For example V might be a dwarf variety. Galton proposes to endeavour by 

 selection to pass back from V to R. If the plants in progress of selection be 

 X, then Galton proposes to pass towards R by selecting the plants above 

 the quartile of X on the R side of the character. He describes very fully 

 how the experiments could be made in an orderly fashion and the needs as 

 to soil and methods of growth; he refers to his paper of April (see our p. 93) 

 for a measure of the rate at which changes might be supposed to take place. 

 No doubt much might be learnt from such experiments — if only, that "retro- 

 grade selection" is impossible. I am not certain that Galton had not this in 

 mind, for if in his view stable varieties could only originate in sports, we 

 could not select back to the original stock, unless selection itself conduces to 

 sporting. I do not think the paper led to any actual experiments on Gal ton's 

 part, although the Editor of the Chronicle wrote strongly in favour of such 

 experiments in a leader in the following week, and there were some sugges- 

 tions on May 29, 1897*. 



In Nature, November 4, 1897 (Vol. lvii, p. 16), Galton gave a brief 

 account of E. T. Brewster's paper on "A Measure of Variability and the 

 Relation of Individual Variation to Specific Differences f ." Brewster is really 

 comparing what we now term Interracial with Intraracial Variability, 

 measuring his variability by what is practically the coefficient of variation 

 V. His thesis is that if for any two interracial characters A and B, 

 V A is >V B , then for the corresponding intraracial characters in the "allied 

 races" v a will be > v b . There does not seem any obvious theoretical reason 

 for this, but Galton holds that Brewster "has provisionally established 

 his thesis that whenever any special character varies much in individuals of 

 the same race, it is probable that it will be found to vary much in 'allied 

 races' and conversely." 



The next paper to be considered is entitled: "Hereditary Colour in 

 Horses," and appeared in Nature, October 21, 1897 (Vol. lvi, pp. 598-9). 

 Galton tabulated his data from material collected and published by " Tron 

 Kirk" in the Chicago journal, the Horseman (Christmas Number, 1896). In 

 the fundamental table which Galton gives there are 3025 matings of bay sires, 

 but as "Tron Kirk" informs us that 3100 foals were born to no more than 

 46 different bay sires J, or an average of 67 foals to the sire, it is clear that 



* The interpretation put by the practical gardeners was that Galton wanted to go back from 

 an improved form to a poor original. But I do not think this by any means the chief purport 

 of his paper; he wanted if possible to go back from a specialised variety to the form, not necessarily 

 inferior, from which it had been obtained. 



f Proceedings, American Academy of Arts and Sciences, May, 1897. 



% It is not said that the matings of bay sires cover these 3100 foals, but presumably they 

 do. The difference in numbers may be due to the omission of grey foals or to twinning. 



