76 CORRELATED VARIATIONS. 



Here we see that the maximum amount of correlation 

 was observed between stature and height of knee, and 

 the minimum between stature and head length. Even 

 in this latter instance, however, the correlation was 

 fairly marked. Thus a constant of .35 indicates that 

 in men 1.75 inch, or 1 unit, above the mean stature, the 



QH 



length of head will on an average be ^-^ X .19 =.0665 



-L . UU 



inch above the mean, .19 inch being the probable error 

 of variation of the head length. The height of knee, 

 on the other hand, would on an average be no less than 



Q 



:p-r X .80 = .72 inch greater. The various medians or 



middlemost values and probable errors found by Gal- 

 ton are as follows : 



MEDIAN. PROBABLE ERROR. 



DIMENSION. INCH. CENTIM. INCH. CBNTIM. 



Head length, 7.62 19.35 .19 .48 



Head breadth, 6.00 15.24 .18 .46 



Stature, 67.20 170.69 1.75 4.44 



Left middle finger, 4.54 11.53 .15 .38 



Left cubit, 18.05 45.70 .56 1.42 



Height of rt. knee, 20.50 52.00 .80 2.03 



In order to determine the degree of correlation be- 

 tween any two organs, it is therefore necessary to adopt 

 the following procedure. Sort out all the individuals 

 into groups such as, for instance, in the case of stature, 

 those varying from 64 to 65, 65 to 66, 66 to 67 inches, 

 and so on, and then determine in each of these groups 

 the mean of all the deviations from the average of the 

 organ of which the correlation with stature is to be 

 determined. Thus, in the group of individuals 64 to 

 65 inches high, the average difference of all the indi- 

 vidual cubit measurements from the median of the 



