REGULARITY OF RETURN 221 



turn. Several of them occur likewise in Cooke's table of average 

 variation, from which his figures are taken (1913). On the basis of 

 the standard deviation, the warblers alone exhibit a variation as small 

 as 10 days or less, and this applies to but two-thirds of the returns. 

 As an argument for an "accuracy of time sense little short of mar- 

 velous," it is unconvincing, and particularly so when viewed in the 

 light of the performance of spring flowers, which possess no "time 

 sense." Incidentally, it is difficult to understand why this sense in 

 warblers should be twice as poor in fall as in spring (cf. page 220). 



Irrespective of other considerations, it is manifest that wide ranges 

 in time may furnish much cogent evidence for control by tempera- 

 ture and food than by length of day as such. The latter is per- 

 fectly regular year after year, and it appears quite difficult to corre- 

 late such a highly variable phenomenon as date of return wuth it. 

 This factor has seemed particularly attractive in the case of species 

 that start south in July, but this explanation has not reckoned with 

 the slow decrease in day length at this time. During the first 2 weeks 

 after the June solstice, this amounts to but 8 minutes for the day, and 

 for the first month only to 17 minutes less of morning and 7 of after- 

 noon. These correspond to decreases of less than 1 and 3 per cent 

 respectively, even the latter appearing quite too small to be tangible 

 in terms of food gathering. By the September equinox, the day is 

 3 hours shorter, but by that time frost is beginning through the 

 United States and surface ice is frequent to the northward. In short, 

 by the time length of day has decreased to the point where it is effec- 

 tive, reduced temperatures are in operation much more decisively, and 

 these then act in combination with food shortage in terms of ability 

 to maintain body warmth. During the spring months of migration, the 

 rate of increase in day length is decidedly more rapid, but even at that 

 it lags behind the mounting temperatures of April and May, which are 

 directly related to the bird's metabolism and bodily comfort. 



In this connection, it is interesting and probably significant that 

 the leaf behavior and the blooming of plants, which are known to be 

 directly connected with temperature, exhibit much the same degree 

 of fluctuation. It is a happy coincidence that the most remarkable 

 record of phenology in America at least was made by Mikesell from 

 1873 to 1912 at AVauseon, Ohio (Smith, 1915), which is not much 

 more than 100 miles from Oberlin. During this period, the time of 

 blooming of 32 out of 39 forbs varied from 20 to 50 days, while that 

 of 46 woody plants ranged from 22 to 57 days. The change of foliage 

 in the autumn was of the same general magnitude, though the average 

 was much higher. As suggested earlier, these results support the view 



