I Il I IV 
BBE Ar En AAG TTG 
REDE oe ete) ci 2 WON, V0 ig Dare ae 0 
165. ETM Sate a ry aL nae 
PEGG) Se Epes kp oe Sater 
AS aha os Bern ean a ER 
[eG si Ae Berens ag" hE Tagua flee eae 
TAO ERE AE 
ISnO her ALE OR vB ES 
IS pdre OG) ht Oe VE 
SRD or DO. NB Es TD De ED 
187: NEI MEDE Me 
The results of the four computations are nearly accordant. The 
value of the temperature coefficient appears to have varied far less 
than it did subsequently. A small fluctuation however, of the same 
nature as that which existed afterwards, appears to have occurred. 
It might be allowable to assume, in accordance with the second com- 
putation, which in my opinion is to be preferred: 
1863—66 Ac= aL 9) C = —.03:0165 
iSo7— 71 Zy) EOS 
187273 ope — 0.0148 
From all the years together we should find 
L3G3a— 13 Press = C= = 00175 
The investigation about the existence of a quadratic term I only 
executed for the mean of the 11 years. 
For this purpose I used the deviations according to the second and 
fourth computation. 
If 4c, and c, represent the correction of the coefficient of #—t, and 
the coefficient of (—t,)*, ¢, being the mean temperature (= + 8°.6 R.), 
we have, expressing both in tenthousandth parts of the second as unit, 
A C, Cy 
2nd Comp. + 0.5 — 0.92 
4th on — 6.2 — 0.43 
At least for the mean of the 11 years, therefore, a quadratic term 
must be quite insensible. 
15. It seemed unnecessary to apply corrections to the reduced 
rates I on account of the temperature coefficient, before proceeding to 
the investigation of the supplementary term. For the mean value of 
