REGULARITY OF RETURN 219 



time the chimney swift varied 13 days. However, the irregularity of 

 this species amounted to 23 days in the 25-year record of \\'oods and 

 Tinker at Ann Arbor and 22 days in the 41 -year record of Jones at 

 OberHn. It rose to 34 for the combined record of a half-century at 

 Ann Arbor, and was still greater for the autumn departure, namely, 

 45 days. Cooke (1913) based his conclusion as to variation being 

 small from year to year upon a single record of 6 years. 



Over the period from 1896 to 1930 at Oberlin, Ohio, Jones (1931) 

 has noted 52 common species that vary 20 days or more in the date of 

 arrival in spring, i.e., between the earliest and latest dates for all 

 the years of record, and 24 with a range of 4 weeks or longer, such 

 as cowbird and fox sparrow, 40 days, rusty blackbird and hermit 

 thrush, 39 days, etc. Data compiled by Roberts for southern INIinne- 

 sota (1932) as to the spring return of 50 species gave a variation of 

 17-19 days for six, 3-4 weeks for another six, and 1-3 months for 

 the remainder. The irregularity of return is still more striking in the 

 longer 50-year record of Wood and Tinker for the vicinity of Ann 

 Arbor, Michigan (1934). 



However, it should be pointed out that the above data refer to 

 first records and not actual arrivals and make no allowance for varia- 

 tions in observation. Most migration observation is fragmentary at 

 best, especially over a long period of years, since most observers, both 

 amateur and professional, are not able to be in the field every day. 

 The amount of observation varies from time to time and especially 

 from year to year. Consequently, there is often considerable differ- 

 ence between the "first record" and the actual (unknown) arrival of a 

 species. Also, the migration data are not quantitative. Too much 

 emphasis is placed on the occasional extreme early or late dates in- 

 volving only a few individuals ; rather the appearance of the bulk of 

 the individuals of the species should be considered. As a conse- 

 quence, accurate determination of the variation in mass arrivals can- 

 not be made from even the best long-term data now existing. 



Nevertheless, it would appear that the amount of variation is 

 greater than commonly supposed, especially during early spring and 

 other periods when the weather is less stable. The use of average 

 date of return has given an impression of regularity where proof is 

 wanting and has led to the attitudes justly criticized by Phillips. 

 Lesser regularity is indicated by average deviation and standard 

 deviation, which do give a serviceable statistical expression of 

 irregularity. They must be interpreted in the light of the total 

 range in return, since the standard deviation, for example, means that 

 only two-thirds of the cases fall within the period of twice its length, 



