(269° ) 
Within that period L could dispose of 182 time-determinations 
at average intervals of 6 days, giving 181 values for the daily rate. 
We can assume as mean error of the result of a time-determination, 
largely accounting for systematic errors such as variations of the 
personal errors of the observers, + 0.04. 
I do not give here in full the results of the comparison of these 
181 observed rates with the two formulae and only lay down the 
mean values found in both cases for a difference: observation—com- 
putation. 
1 found: 
Formula | M. Diff. == = 03.0333 
i il + 0 .0344 
Hence this mean difference is nearly the same for the two for- 
mulae; indeed, if the three years are kept apart, it is found to be a 
little greater for formula I in two of the three years. 
We may therefore say that the two are in equally good agreement 
with „the observations and for the investigation of the barometer 
coefficient it was sufficient to use either. 
I chose formula If (linear influence of the temperature) and I 
proceeded in the following way. The rates reduced with that for- 
mula to 760 m.m. and 10° and freed from the supplementary ine- 
quality were divided into five groups according to the barometric 
pressure and for each group the mean of those reduced rates was 
calculated. The results are laid down in the following table, where 
the first column gives the number of rates from which each mean 
has been derived. 
Number. | Barom. Reduced. D. R. 0.—C. 
17 132.8 — 0:.174 — 0s .002 
3H | 757.6 | 162 ay 
GS 762.6 | 154 ee OH 
4h 767.4 145 send 
| 7 , 
21 | 112.2 | 141 — 02 
From these results [ derived as correction for the barometer 
coefficient: . 
Ab = + 0:.0017 
joe 
