THE MEASUREMENT OF VARIATION. 25 



Here we see that most of the face measurements are 

 far more variable than most of the head and limb 

 measurements; that of the nose height, for instance, 

 being six times as great as that of the head length. 

 The high value which is universally accorded to facial 

 proportions as a means of personal identification thus 

 receives its numerical justification. 



On comparing the variability of the measurements 

 in the individuals of eight different races, it was found 

 to be more or less the same in each case. If the nose 

 of a Jew is a very variable organ, so is that of a Slav, a 

 Magyar, or a Chinaman. 



Davenport and Bullard * used the method of arith- 

 metic mean error in the 4000 enumerations which they 

 made of the Mullerian glands in the forelegs of swine. 

 These glands vary in number from to 10, the average 

 being 3.53. The arithmetic mean error was 1.41 in 

 male swine, and 1.38 in female swine, or the variability 

 was 2.5 per cent, greater in the one case than the other. 

 Again Garstang f has used it to estimate the variability 

 of various local races of the mackerel. 



There is still another method of estimating varia- 

 bility, which is more accurate than either of the two 

 mentioned, but which until recently has not been used 

 so frequently as they were, because of the labour of ap- 

 plying it. This is the method of Mean or Least Squares. 

 One determines the deviations from the average in the 

 same way as for the arithmetic mean error, but then 

 squares each of them, takes the sum of these squares, 



*Proc. Amer. Acad. Arts Sci., xxxii. p. 87, 1896. 

 f Jour. Marine Biol. Asso., vol. v. p. 235, 1898. 



