THE MEASUREMENT OF VARIATION. 



19 



called the " Probable Error ' ' of variation. In a per- 

 fectly normal curve, these values are equal in value 

 and opposite in sign, but as no experimental result is 

 perfect, they usually differ slightly in amount. A 

 mean between the two is therefore taken, and this is 

 denoted by the symbol Q. 



For the practical determination of the probable 

 error, however, it is quite unnecessary to plot out the 

 results in the form of a curve. The method adopted 

 is best illustrated by a concrete instance. In the ac- 

 companying table are given the results obtained by 

 Mr. Galton * for the strength of pull, as of an archer 

 with a bow, of 519 males, aged 23 to 26 : 



Here we see that the numbers of actual cases in each 

 group are given in the second column, and that they are 

 calculated as percentages in the third column. They 

 are summed from the beginning in the fourth column, 

 and we thereby gather that whilst only 37 per cent, of 

 all the men had a strength of pull under 70 Ibs., 70 per 

 cent, of them had one under 80 Ibs. It can be calcu- 



*" Natural Inheritance," p. 199. 



