( 276°) 
rc 
riority of Honwt 17 in its present state. That this regularity markedly 
surpasses the one reached formerly is shown also by the results of an 
investigation of the vears 1886—87, which are among those of the 
ereatest regularity in the period 1877—1898. This investigation was 
made in a similar manner as the present one, the mean interval 
between the time determinations used was 5 days and the mean 
error found was = 0+.0365. 
6. We may also investigate the rates of a clock in such a manner 
‘that only the irregularities of a very short period are considered. 
A simple process for attaining this is to caleulate the mean value of 
the difference between two consecutive reduced daily rates. 
Applying this method to Honwt 17 during the period under 
consideration *) I found: 
Mean difference 1899 May—1902 Sept. + 08.0315. 
; 7 1901 May 1902 Sept. + 0%.0253. 
From these mean values considered in connection with the mean 
errors of the rates in 6-daily and in monthly intervals formerly found 
we can draw some, albeit rough, conclusions about the amount of 
the perturbations of longer and shorter periods. 
The values found, as well those for the whole period as those 
derived for the last year only, are given in the following table. The 
columns A contain the values found directly, the columns / those 
diminished by the amount that can be ascribed to the errors of 
observation, assuming + 004 as the total mean error of a time- 
determination. M.E. @ of a 6-daily rate stands for the total mean 
difference from the formula Iv, found above, M. B. « represents the 
error derived from the mean differences between two consecutive 
rates. The mean errors of the monthly rates differ a little from 
those of my previous paper as they now also refer to formula Ia. 
‘mmm 
1899 —1902. 1901—1902, 
M. Diff. of two 6 d.r. | + 050313 | = 050267 | = 080253 | = 0.0193 
M. E. z of 6 d. r. | 189 | | 137 
| | 
M. E. # of 6 d. r. | 314 | 207 251 | 233 
M. E. of monthl. r. | 209 208 164 163 
1) The rates were reduced by means of formula Ila, but a reduction according 
to Ia would practically have led to the same result. 
