( 275 ) 
A consideration of the differences obs. [Ie and obs. shows 
however, that the latter conclusion does not equally hold good for 
all parts of the year and that the agreement with formula Hd is 
especially bad in the winter months. In order to investigate this more 
closely, I divided the observations into groups of two months and 
calculated for each group the mean value of the differences, first for 
each year separately, then after combining the corresponding groups 
of the different years. The latter values follow here. 
Form. Ha. | Form. Ib. 
Jamaar y,sKebPUarg: na aie cues eae 2E 0s.0402 | =F Qs. 0549 
RPE AEN se aa ale OEREN ed 208 214 
DEAN UNE Ber et Alant oe vei ken 285 284 
TIN AATIETSG. lor sn tie BBL mend 423 368 
September, October........ oden 215 | 932, 
November, December ......%....... 309 | 559 
They lead to the singukw result that during the four winter months 
formula Il / agrees much less with the observations than lr, whereas 
in the middle of the summer the agreement with 115 seems to 
be better, and in the other months both formulae may be said to 
agree equally well. In this respect the different years practically 
lead to the same conclusion and hence we cannot say that this has 
been brought on by entirely accidental causes. However this may be, 
we are not entitled yet to assume a lagging behind of the influence 
of the temperature. 
Let us now consider separately the resuits for formula Ta, which 
seems to represent the observations with the greatest precision (those 
for Ila do not essentially differ from them). It will be seen immediately 
that during the last seventeen months the rate has been considerably 
more regular than during the first two years *); a smaller M.E. has 
been reached although the 5 last of these 17 months were not 
included for the derivation of the formula. Thus the feature observed 
before, i.e. the gradual improvement of the regularity of the rate after 
the mounting of the clock, shows itself once more. The mean result 
for the whole period (M.E. — + 08.0811) may already be regarded 
as very satisfactory, and the great regularity represented by a mean 
difference of + 08.0251 between a daily rate from a 6 days interval 
and a relatively simple formula gives us a high sense of the supe- 
1) Already at the beginning we had left out the first 4 months after the remounting. 
