3ü2 



(hl thc Lau'S qf f uheritanre in Man 



regression lines of the 78 correlation tables, but thc t'oll(]\viiig three cases arc a 

 fair random sample cf wliat aotually occiirs*. 



Diagrain I. Stature of Father and Son. 

 Diagrain II. Spcni in Mother and Daughter. 

 Diagrain III. Brothers Forearni and Sisters Span, 



thc latter beiug an cxaniple of a cross-correiation. 



DiAGii.VM I. l'robiMc Stature of Soll for given Father't Stature. 

 Eegression Line: 5=33-73 + -516 F. 1078 Cases. 



6S ao ei 63 63 64 es 6e e? es eo 70 71 72 73 74 7S 



Father's Stature ( = F) in itiches. 



It will be Seen fruni thcse cases that, o.xcept lU'ar the terniinals, whcre thc 

 nunibers of cases are very few, that the regression is ciosely linear. We are thus 

 rclicvcd from any dilticulties about regression or correlation. We have only to 

 find the ordinary coefficient of correlation r, and the regression coefficient ra^ja^, 

 and thcsc will sutfice to describo tbe average degree of hereditary rcscrnblance. 

 All this is done without any assnmption of the normal curve of frequency. As a 

 matter of fact, however, the normal curve very ciosely siiffices to dcscribe the 

 distribution of niany physical characters in a human population. This is illus- 

 trated in the acconipanying diagrams which are fair samplos of stature and 

 span frequencies. In Diagram IV. we have the foUowing data for stature in 

 mothers, plotting frequency observed against theoretical frequency. 



• A furtlier case from tlic data, Hiat of cubit in Fatlicr aiid Son, was given in liiometrika, Vol. 11. 

 p. 'iK;. 



