Physical Significance of the Mean Diurnal Curve of Tenijjerature. 189 
will be higher or lower according as t is higher or lower. We have then 
to find the most common value of t - f at any hour. For if the mean 
diurnal curve of temperature has a real physical significance the most 
common successive hourly values of t — t' should give the same curve. 
On account of the great labour of the inquiry it has only been possible 
to prosecute it for the months of January (as representing the summer) 
and July (as representing the winter). But for these two months it has 
been done for ten whole years of hourly temperature readings. As 
an illustrative example of the method the foregoing table will perhaps 
be sufficient. 
Here the first column gives the dates ; the second and third columns 
the Kimberley temperatures at 3 p.m. and 4 p.m. ; the fourth column the 
differences, counted positive when t is the greater. When each July 
of the ten years 1902-11 has been treated in this way, particular values 
of t - t' are counted and the total frequencies arranged in rows in the 
following way : — 
July XV.-XVI. 
t - t'. 
No. 
t - t'. 
No. 
0 
Less than 0*0 
24 
1-3 
8 
0-0° 
14 
1-4 
5 
0-1 
10 
1-5 
14 
0-2 
6 
1-6 
5 
0-3 
15 
1-7 
7 
0-4 
18 
1-8 
10 
0-5 
11 
1-9 
6 
0-6 
17 
2-0 
5 
0-7 
14 
2-1 
4 
0-8 
19 
2-2 
6 
0-9 
16 
2-3 
4 
1-0 
20 
2-4 
4 
1-1 
14 
Greater than 2-4° 
20 
1-2 
14 
That is to say, in the ten years 1902-11, between the hours 3 p.m. and 
4 p.m., the temperature fell, for example, by 0°-3, 15 times ; by 0°-9, 
16 times ; by 2°-0, 5 times, and so on. In the Tables at the end will 
be found the total frequencies for each hour, arranged in sets of five 
consecutive differences. 
A remarkable feature of the results given in these Tables is the fact that 
over a wide range of temperature differences there is no strong central 
value about which the frequencies cluster. It is, in fact, not possible to 
