THE MEASUREMENT OF VARIATION. 35 



slightly asymmetrical curves of variation, the varia- 

 bility of the bilateral homologous measurements being 

 always, with one exception, slightly higher on the blind 

 than on the eye side of the fish. Still the departure of 

 some of the curves from the normal Gaussian curve was 

 only very slight. The degree of difference between the 

 actual frequencies obtained in any series of measure- 

 ments and the theoretical frequencies calculated from 

 the type of curve found to show the closest agreement 

 with them, is best represented by determining the per- 

 centage difference of each actual frequency from each 

 theoretical frequency, and then taking an (arithmetical) 

 average of the whole. This average percentage differ- 

 ence of theoretical and actual values may be repre- 

 sented by the sign A . In the case of four of the above 

 series of measurements, the A was only 2.20 per cent, 

 when the measurements were compared with a normal 

 Gaussian curve, and 1.82 per cent, when compared with 

 a slightly asymmetrical curve (Pearson's Type IV). 

 The " fit " was therefore better with the asymmetrical 

 curve, but only very slightly so. It is open to question, 

 therefore, whether any practical object is gained by 

 estimating exceedingly slight degrees of asymmetry in 

 series of measurements. The labour of so doing is very 

 considerable, and it may well be doubted whether it 

 would not be more profitably employed in making more 

 extended series of observations, and subjecting them to 

 less rigid examination. Some recent observations of 

 Miss Hefferan * are instructive in this connection. 

 These were made upon the frequency of distribution of 

 the numbers of teeth on the jaw of an annelid, Nereis 

 * Biol. Bulletin, vol. ii. p. 129, 1900. 



