(OF } 
These means have further been used instead of the original values. 
5. The reduced daily rates 1, as given in the tables above, formed 
the basis of the further investigation. The first 5 months have been 
excluded from the beginning. 
To begin with, it is possible without much computation, simply 
by combining the reduced rates into groups, to show that they must 
still contain a term of yearly period which cannot be explained by 
a direct influence of the temperature. 
This is done as follows. The monthly means of the rates and of 
the temperatures were arranged in groups of one year each, the year 
beginning with May and ending with the following April. Then the 
means were taken of the rates for each year, and the differences between 
the monthly means and their yearly mean were formed. Thus I derived 
for each year a series of 12 differences: monthly means of rate — 
yearly mean, and also a series of 12 corresponding temperatures. In 
each of these series the mean was then taken of the first and the 
last value, of the second and the last but one, and so on. Finally also 
similar results were derived substituting for the temperatures the 
differences between the actual temperature of the month and that of 
the preceding month (4 Temp). 
Then the same process was repeated with the only difference that 
the vearly groups commenced with February and ended with January. 
The aim of this process will become clear when the results are 
considered. For brevity’s sake I confine myself here to the five years 
1884 to 1888. The differences: monthly means of rate — yearly mean 
are given separately for each year, and also for the mean of the five 
years. For the temperatures and the 4 Temp. only the means are given. 
The differences of the rates are expressed in thousandth parts of a 
second as unit. 
It is evident at first sight that in the first arrangement all the series of 
rates show a very marked progression, while the temperatures are 
nearly the same. In the second arrangement the reverse is true. On the 
other hand the variation of the rate is roughly parallel to that of 
the 4 Temp. Hence the rates contain a term which does not depend 
on the actual temperature, but of which the maxima coincide with those 
of the yearly change of temperature, or, in other words, the yearly 
periodicity in the rate cf the clock does not coimeide with that of 
the temperature, and from the values just quoted we easily derive 
that the first lags about half a month behind the latter. 
6. Before a closer investigation of the phenomenon is possible, it 
