382 Transactions of the Boyal Society of South Africa. 
These results are to be regarded as approximately correct to the limits 
of accm-acy of spirit thermometers. The readings of the thermometers 
are, if anything, rather higher than the truth on account of the sluggish- 
ness of the instruments ; but a test of their performances as compared 
with mercurial thermometers has given no material support to Walter's 
claim that spirit thermometers are altogether unsatisfactory.* In 
any case the mean differences, D, are not likely to be widely different 
from what they would have been if mercurial thermometers had been 
used. 
The differences between the minima, given in the Table, are not strictly 
comparable, because there is nothing to show that the minima registered 
by the two thermometers fall exactly at the same instant of time. It is, 
however, certain that the synchronous differences average high, especially 
in the winter months, when a value for D of 10° F. is by no means 
uncommon. 
The interpretation of the numbers given in Table 1 is not at first sight 
very obvious. It can be seen upon following the run of the observations 
day by day that the results depend in some way upon the state of the sky 
and the moisture conditions of the atmosphere. On cloudy nights, that 
is to say, or on clear nights after rain, the radiation-temperature gradient, 
K . dO/dx say, measured upwards, will mostly be very small, and occasion- 
ally even reversed in sign. Yet, on the other hand, in cloudy and rainy 
weather considerable gradients are often observed. 
Therefore, for the purpose of getting into a better position for the 
interpretation of the results of Table 1, a series of hourly observations 
of the simultaneous readings of the two thermometers between the 
times of sunset and midnight was commenced in May, 1910, and con- 
tinued until August, 1911. The mean results for this piece of work 
are given in three sets in Tables 2, 3, 4, for the five hours 8 p.m. to 
midnight, i.e. — 
1. Table 2, for very clouded skies in which the cloudiness exceeded 
50 per cent, for the whole evening ; 
2. Table 3, for skies less than one-half obscured by cloud ; 
3. Table 4, for unclouded skies. In this case the refinement of 
distinguishing degrees of clearness was attempted but not 
carried out. 
* A, Walter: " Discussion of the Errors of certain Types of Minimum Spirit Ther- 
mometers in use at the Eoyal Alfred Observatory, Mauritius " {Trans. S.A. Phil. Soc, 
1907). 
