2 'The Wilson Bulletin — No. 82. 



of exceptionally warm weather may induce the birds to ar- 

 rive somewhat earlier than usual. But these variations from 

 week to week largely balance one another and the average 

 dates of arrival for the whole season show from year to year 

 very slight variations. 



Thus at Lanesboro, Alinn., the average date of arrival for 

 the ten years, 1884-1893, is 144.8 days after January 1 ; the 

 variations are: 1886, 0.2 days late; 1887, 0.3 days early; 

 1888, 0.7 early; 1889, 0.4 early; 1890, 1.2 days late. The 

 average variation is 0.6 days, and the maximum variation^ 

 1.9 days, from 0.7 days early to 1.2 days late. The year 

 1885 is 1.9 days late, but there are evidences in the notes 

 that this is due to lack of time in the field on the part of the 

 observer rather than to actual variation by the birds. 



The same test was made of the migration dates from Keo- 

 kuk, Iowa. The year 1892 is 1.0 day earlier than the aver- 

 age ; 1893, 0.1 day later; 1894, 1.7 days late; 1895, 0.8 days 

 early ; 1899, 0.5 days early ; 1900, 0.5 days late. The ex- 

 tremes of variation are 2.7 days and the average variation 

 0.9 days. 



Grinnell, Iowa, shows much the same : 1885, 0.7 days late ; 

 1886, 0.2 days late; 1887, 0.7 early; 1888, 0.6 early; 1889, 

 0.1 late; 1890, 1.0 late; average variation only 0.5 days; max- 

 imum variation, 1.7 days from 0.7 days early to 1.0 day late. 



The averages for Aweme, Manitoba, are: 1898, 1.2 days 

 late; 1900, 0.2 days late; 1901, 0.8 early; 1902, 0.8 early; 

 1903, 0.2 days late ; average variation 0.6 days ; maximum 

 variation, 2.0 days from 0.8 days early to 1.2 days late. The 

 records for 1899 are omitted because that year the observer 

 could not devote the usual time to the work and the average 

 is several days late. 



The records just given include the work of four observers 

 in four widely separated towns during fifteen different years 

 and hence ought to be a fair test. The remarkable agree- 

 ment of the results is a strong indication that they represent 

 the actual movements of the birds. The twenty-one series of 

 observations give a maximum range of 2.5 days from O.S 



