( 208 ) 
13. In the first place I have investigated in how far the non-pe- 
riodie part of the rate, the constant @, has varied during the period 
under consideration. 
For this purpose I have combined the monthly means to yearly 
means. They are as follows, the years beginning with May. 
1862 — 03.316 1868 — 05.400 
1865 309 1869 384 
1864 „350 1870 368 
1865 A21 1871 367 
1866 436 1872 000 
1867 428 1875 98 
It is seen that the negative rate has somewhat increased in the 
beginning and somewhat decreased afterwards and that it remained 
nearly constant during the last four years. 
With these values and the corresponding ones for years beginning 
with August, November and February, I drew, in the same way as 
was done for the period formerly considered, a curve representing 
in a first approximation the change of @ with the time. 
14. In the second place the influence of the temperature was 
investigated. I tried to find out: 
ist. In how far, if we assume a linear influence of the temper- 
ature, the adopted temperature-coefficient applies for the whole of 
the period ; 
Ind whether there is any term varying as the square of the 
temperature. 
For the first investigation the several years were kept separate. 
They were assumed to begin with February. 
I used 1st. the deviations of the monthly means from their yearly 
mean, 2"¢. the deviations of these same monthly means from the 
values of a taken from the curve. In the third and fourth place 
the computations were repeated using, not the monthly means them- 
selves, but the mean value for the first month combined with the last, 
that for the second combined with the last but one, ete. By this device 
the influence of the “supplementary term” must be nearly completely 
eliminated at the outset. 
In this way I found for the correction of the adopted coefficient 
—-0:.0174, the following four series of values; they are expressed 
in tenthousandth parts of a second. 
