752 



SCIENCE 



[N. S. Vol. XXXVI. No. 935 



Alar extent 



Length of head 



Length of humerus 



Length of femur 



Length of tibio-tarsus 



Width of skull /oo 



Length of sternum 



Average 



/3ff «« IMS vm mi /*30 AtM- /47S 



Fig. 5. The survival polygons of the measurements expressed in percentage of 

 length in the Bumpus sparrow ease. Dash line, all males ; dotted line, females ; 

 dot and dash line, young males. The horizontal line opposite the name of each 

 dimension represents 50 per cent, survival, the next horizontal line below and the 

 next above 100. Numbers at bottom are the average dimensions of the four classes. 



(divided to the nearest class) surviving. 

 Bumpus had based his conclusions on the very 

 few extreme individuals, instead of the two ex- 

 treme fourths. I believe this is dangerous, on 

 account of the small numbers, but for com- 

 parison, I have also plotted the extreme classes. 

 To avoid the danger of small numbers, I have 

 combined the extremes in all six character- 

 istics, since the survival curves have a com- 

 mon inclination. Here also the result (Fig. 

 6) shows secular and not periodic selection. 



Bumpus was led to his conclusion from the 

 fact that in length there is some degree of 

 periodic selection as well as secular selection 

 shown (Pig. 5). Even in length the small 

 numbers involved in the end classes cast some 

 doubt as to their significance. Taken as a 

 whole, therefore, I believe this sparrow catas- 

 trophe reveals very much more secular than 

 periodic selection, if there be any of the latter, 

 although very naturally the experiment has 

 been quoted to show periodic selection in the 



