Vol. 5 No. 2 



TORREYA uBRA.v 



NEW \ u:»K 

 February, 1905 BOTANUaL 



(JARiJii.N 



GALTONIAN REGRESSION IX THE "PURE LINE"* 



By Gicorck Harrison Shiii.i, 



Among the experiments undertaken this year at the Station 

 for Experimental Involution for the purpose of investigating the 

 inheritance of characters in plants, was one intended to be essen- 

 tially a repetition of Joliannsen's studies f in the inheritance of 

 seed-weights in beans. The variety of PJiascolus vulgaris chosen 

 for this study proved to be unsatisfactory from a technical stand- 

 point and it is not proposed to pursue the experiment further 

 with this material, though several subsidiary questions may be 

 taken up in other plants. The relation between the results of 

 Johannscn on beans and those of Galton on sweet-peas X have 

 appeared on further analysis to be in need of reinterpretation 

 rather than reinvestigation, and the writer feels justified, there- 

 fore, in taking this abandoned experiment as a text for such re- 

 interpretation. 



From a number of statistical studies upon various characters 

 in man and animals and a single series of experiments in 

 sweet-peas, Galton derived his law of natural inheritance and 

 its corollary — the law of regression from mediocrity. || The 

 law of natural inheritance is, briefly, that the offspring of any 



* Presented before Section G, A. A. A. S., at Philadelphia, December 30, 1904, 

 under title of " Inheritance in Pure Lines." 



f Ueber Erblichkeit in Populationen und in reinen Linien. Jena : Fischer, 1903. 



J Natural inheritance. New York : Macmillan & Co., 18S9. 



II This has frequently been called " regression tcnoard mediocrity," but as the co- 

 efficient of regression is measured y)('w the mean condition of the population confusion 

 has arisen through expressing it in this way. Galton's own inconsistency in discus- 

 sions of regression is doubtless responsible for this confusion. He first presents it 

 clearly as a deviation from mediocrity, but later says there is "no regression at all " 

 when this deviation is etjual in the two kinships under comparison, and the coetficient 

 of regression is unity. Cf. Natural inheritance 95-9S with 132-133.) 

 [Vol. 5, No. I, of ToRRKYA, comprising pages I-20, was issued January 19, 1905.] 



21 



