NATURAL SELECTION 209 



elytra of the beetle Diabrotica soror over the period 1895- 1902. 

 The difference consisted in the dominance in 1 901 -1902 of a 

 modal condition which was not dominant in 1895. 



Beyond stating that it is not likely that the change in 

 position of the spots or the elytra would serve as a basis for 

 selection, the authors produce no evidence that the change 

 is not due to selection. 



(16) Bumpus (i8gg) : alleged selective elimination in Passer 

 domesticus. 



' After a severe storm of snow, rain and sleet a number of 

 English Sparrows were brought to the Anatomical Laboratory 

 of Brown University. Seventy-two of these birds revived : 

 sixty-four perished.' It was the purpose of Bumpus's study 

 to show that the birds which perished did not die from accident 

 but because they were physically disqualified, and that the 

 survivors lived because they possessed ' certain physical 

 characters ' which enabled them to withstand a particular 

 phase of selective elimination. He measured 9 characters 

 (e.g. length, weight, alar extent, etc.) of the dead and survivors. 

 He divided his specimens according as they were adult or young 

 and male or female. He found that there were differences 

 in some characters as between survivors and eliminated and 

 not in others, and he assumed (p. 213) that there were funda- 

 mental differences between the dead and the survivors. As 

 the numbers in each group thus discriminated are low (the 

 total which died was only 64, of which 24 were adult $ and 

 12 were young o*)> and as he compared the averages of the 

 various groups, it will strike the modern statistical biologist 

 that his conclusion is premature. These observations, sug- 

 gesting a selective elimination, have been widely cited as 

 proving the general occurrence of such elimination. 



Harris (191 1), however, on the very full data published 

 by Bumpus, produced the necessary statistical constants 

 (standard deviation, etc.) and applied the usual tests for 

 significance. His treatment of the subject is rather peculiar. 

 He admitted that, by applying the usual statistical tests, 

 differences of a statistical value varying from ' significant ' 

 to ' possibly significant ' were actually to be obtained from 

 Bumpus's figures for some (but by no means all) of the charac- 

 ters. Yet he concludes that, ' though the cautious biometrician 



