BLASTOGENIC VARIATIONS. 127 



of the general population as theirs did. Thus, if the 

 mid-parental stature (the average between the statures 

 of the man and the transmuted stature of the woman) 

 be 72 inches, or 3f inches greater than the mean stature 

 of the whole population, then the average stature of 

 their sons will be only 3f X f = 2^ inches greater than 

 the mean. If the mid-parental stature be 66 inches, or 

 2^ inches less than the mean stature, then the average 

 filial stature will be 66f inches, or only 1^ inches less 

 than the general mean stature. In addition to calculat- 

 ing the regression between parents and sons, and grand- 

 parents and grandsons, Galton calculated it for col- 

 lateral relationships, as between uncles and nephews, 

 and brothers and brothers. 



Many of the data recorded by Galton in his " Natural 

 Inheritance ? were worked over again by Pearson in 

 his memoir on " Regression, Heredity, and Pan- 

 mixia," * and various improvements in statistical 

 methods suggested. The mathematical measure of re- 

 gression, or coefficient of regression, he defined to be 

 " the ratio of mean deviation of offspring of selected 

 parents from the mean of all the offspring to the devia- 

 tion of the selected parents from the mean of all the 

 parents." It is to be noticed that according to this 

 definition, the deviation of the offspring ought to be 

 measured from the mean of the offspring of the general 

 population, and not of the whole population, both 

 parents and offspring, for thereby factors such as secu- 

 lar natural selection and reproductive selection are 

 allowed for. 



In this memoir, when discussing coefficients of re- 



* Phil. Trans. 1896, A. p. 253. 



