( 204 ) 



the former in their general course. The length of the perioil with 

 which he starts thus amounts to about 423 days and from the 

 observations a correctiou is found for it of -j- o days, which however, 

 as Chandler observes, must be quite uncertain, it not being sure 

 that the length of the yearly period is exadbj a year. Meanwhile, 

 later on, a correction of + 4 days for the length of the period is 

 assumed beside such a one of -}- 8 days for the mean epoch and, 

 as Chandler thinks it proved that the length of the period is 

 variable, he accounts for the correction by a quadratic term added 

 to the formula of the epochs which thus becomes: 



T=: 2412640'! + .127'i.0 7i/— 0-1.08 7^^ (3) 



where the initial epoch is placed 24 periods later than that of the 

 preceding formula. 



Tested by the older observations this formula proved to satisfy 

 fairly those since 1835, but not at all those of Pond, which leave 

 for the epoch a deviation of 166 days. Although formerly Chandler 

 set great store by Pokd's observations, it yet seems that he desires 

 io have the elements of formula (3) regarded as „the revised elements" 

 he wished to determine. It is true that a doubt about this conclu- 

 sion arises by reading in the „conclusions" which, in another part 

 of the paper (p. 107), are derived from „substantially all the competent 

 „testimony available" {b) „that the mean period since 182.5 is 428 days 

 „within a small fraction of a day", whilst formula (3) gives us for 

 this quantity 43P'.G, and {d) that the hypothesis of a change in the 

 period uniform with the time is incompatible with the observations 

 before 1860, whilst in conclusion (e) a change per saltum between 

 1830 and 1860 is called also incompatible with the facts. Leaving 

 this for what it is, I shall in what follows, indicate formula (3) as 

 Chandler 1898. 



The differences between the epochs computed according to this 

 formula aud to that of 1894 are rather small between 1870 and 

 1894, but increase rapi^lly beyond these limits. So we find for 

 Ch 98— Ch 94 in 1830^— i2G d., in ISGO + 38 d., in 1808 

 + 25 d. and in 1900 + 32 d. 



4. In the first place I investigated more closely what the obser- 

 vations from 1890 — 97, taken by themselves, can teach us about 

 the length of the 14-miOuthly period. In my former paper I examined 

 the X of the three last years only; now I did the same for the tliree 

 first years and then I acted in the same way for the //. 



1 thus obtained the following rosalts for the mean epochs of 



i 



