Gravity in South Africa. 
87 
Two thermometers were provided, of which one (No. 53) was used 
throughout. It had been standardised in 1894 and again in 1895. It 
was expected that no further change could occur after the first few years, 
and unfortunately the author neglected to standardise it again till August, 
1910, when it was found to have changed very much (see table, p. 90). 
It was tested by a thermometer (P.T.E. 28,629) possessing the 0° and 100° 
points, and these were found correct. The results given below are 
according to the 1910 standarisation. 
{d) There is no difficulty in finding the density of the air with 
sufficient accuracy. This correction depends, therefore, only on the 
accuracy of the formula given by Sterneck, according to which the 
period varies by an amount proportional to the change of density. For 
corrections c and cl the values obtained in Austria have been 
used, viz. : — 
49 '26 units (sidereal time) per 1° C. 
542-0 ,, ,, for a change of density equal to 
the density of air at normal temperature and pressure. The pressure 
correction might be eliminated by swinging the pendulum in vacuo, did 
the apparatus permit. 
(e) This correction is by far the most difficult. It consists of two 
parts — the oscillation of the pendulum support, and that of the pillar 
or wall on which the support is placed. With regard to the former, it is 
assumed that the correction is always the same, and therefore disappears 
in comparative measurements. The stability of the pillar or wall is 
tested by giving it ten properly timed blows with a wooden rod which 
has been pressed against a spring with a force of 5 kilogs. If the 
succession of blows does not induce a measureable amplitude of swing 
in the pendulum it is assumed that the effect of vibration of the pillar 
is negligible. More satisfactory methods have been worked out of 
late, but they involve the use of different patterns of apparatus (stands 
carrying more than one pendulum at a time), and so could not be used. 
From this review of the corrections it might be supposed that when 
working under favourable conditions the results should agree to within 
5 units. Experience shows that this is not the case. Even at a well- 
equipped observatory, such as that of Pola, differences of 20 units 
are often noticed, and on repeating the observations after an interval, 
much more. The same is true of the South African measurements — 
a fact which long puzzled and disconcerted the present writer. The 
larger differences may be due to secular changes of length in the 
pendulums, but this would not account for the irregular variations from 
day to day. Eventually a cause for this was found in variations in 
* Veroff. Hydrog. Amts. der K. K. Kriegsmarine, No. 7, p. 1. Pola, 1898. 
