3 14 STATISTICAL STUDY OF INHERITANCE 



4. Filial Regression 



It has often been remarked that the children of extraordinarily 

 gifted parents are sometimes very ordinary individuals, and that 

 the children of under-average parents sometimes turn out sur- 

 prisingly well, both physically and mentally. Every one who 

 has looked into the facts of inheritance in greater detail, and 

 has compared the average of qualities in successive generations, 

 has noticed in a general way that there is a tendency to sustain 

 the same average level from generation to generation. Even 

 the older inquirers, like Lucas, called attention to the fact that 

 extraordinary qualities in families tend to wane away, as if 

 there were some mysterious succession-tax levied on marked 

 deviations from the average, whether in the way of excellence or 

 of defect. But we owe to Mr. Francis Galton's careful statistical 

 work the generalisation known as the Law of Filial Regression, 

 which has replaced a vague impression by a definite formula. 

 He has defined and measured that tendency towards mediocrity 

 that tendency to approximate to the mean or average of the 

 stock, which is expressed by the term Filial Regression. We 

 may notice at the outset that this has nothing to do with 

 reversion or with degeneration, that it works upwards as well 

 as downwards, forwards as well as backwards. 



The data which Galton utilised were chiefly the Records of 

 Family Faculties, obtained from about one hundred and fifty 

 families, and dealing especially with stature, eye-colour, temper, 

 artistic faculty, and some forms of disease. These were supple- 

 mented by measurements at Galton's anthropometric laboratory, 

 and by observations on sweet peas and to some extent on moths. 



Most trustworthy, however, were the data procured in regard 

 to stature, which, as Galton points out, is a quality with many 

 advantages as a subject of investigation. It is nearly constant 

 during mature life, it is readily and frequently measured with 

 accuracy, and it does not seem to be of appreciable moment in 



t 



