CERTAIN LAWS OF VARIATION. 213 



breadth, general shape, and colour. The American 

 eggs were slightly more variable in length, they vary- 

 ing from 18 to 26 mm. instead of from 18.5 to 25 mm., 

 and also were reduced in their average length from 22 

 to 21 mm. As Davenport has shown,* however, the 

 mean deviation or arithmetical mean error, which is a 

 better index of the variability, is somewhat smaller in 

 the American eggs than in the English; but the arith- 

 metical error of the variations in the relation of breadth 

 of egg to length is slightly greater in the American eggs 

 than in the English (in the proportion of .73 to .70). 

 In order to compare the variations in the shape of the 

 eggs, such as conical, ellipsoidal, pear and lemon shapes, 

 Bumpus requested a disinterested person to select from 

 the 1736 mixed eggs the hundred eggs which appeared 

 to him most variable. The American eggs were known 

 by a secret mark, and it was found that 81 of the 

 selected eggs were American, and only^!9 English. By 

 a somewhat similar process of selection with reference 

 to extremes of colour marking, it was found that 82 of 

 the 100 selected eggs were American, and 18 English. 

 There can be little doubt, therefore, that American eggs 

 are considerably more variable, both in shape and 

 colour, than English eggs. 



The conclusions to be drawn from this most interest- 

 ing series of observations, however, are probably not 

 those suggested by the author. In the first place, as 

 Davenport has pointed out, they afford presumptive evi- 

 dence that the American eggs, or the birds which pro- 

 duced them, were subjected to a distinct selective 

 process. Thus the curve of distribution of the length 

 * L'Annee Biologique, 1897, p. 496. 



