A.—MATHEMATICAL AND PHYSICAL SCIENCES 35 
decrease in unemployment ; so that the last two factors varied similarly. 
We see, then, that we may be misled if we do not take into account all the 
factors that may be operative. In other words, statistical methods like 
logarithm tables are invaluable as a tool for giving correct numerical 
results with the minimum of mental labour ; but neither tool possesses 
imagination or judgment, and neither of them is a substitute for expert 
knowledge of the subject to which it is applied. 
Let us now turn to the North Pacific Ocean which, in spite of its 
limited access to the Arctic seas, is subject to fluctuations very similar 
to those of the North Atlantic. A similar treatment yields Fig. 5, in 
which increased pressure gradients go with high temperature to the 
north-east and south-west, and low temperature to the north-west and 
south-east. It will be noted that in both the North Atlantic and Pacific 
1880-( 1890- 91 1910-11 1930-31 
FOLLOYVING 
SUMMER, 
THE SOUTHERN OSCILLATION. 
Fic. 8.—Forecast of December to February from previous June to August. 
Oceans a fluctuation is classed as positive when the pressure gradient 
is strong and the wind circulation is active. 
The largest known system of related seasonal weather is that called 
the ‘ southern oscillation ’ (or ‘ southern fluctuation ’), which has features 
in the southern summer of December to February somewhat different 
_ from those of the southern winter of June to August. It will be seen 
in Figs. 6 and 7 that at both times of the year the fluctuation is called 
positive when pressure is high in the southern Pacific and low in the 
‘Indian Ocean, and temperature is mostly low in the Tropics ; but. the 
economic importance is in connection with rainfall, for the fluctuation 
has a correlation coefficient of over 0-8 with the summer rainfall of north- 
east Australia, over 0-7 with the monsoon rainfall of India and with the 
Nile floods, 0:6 with the rainfall of large areas in South America, and 
over 0°5 with that of a region in South Africa. 
A surprising fact comes out on comparing the numerical series giving 
the characteristics of the summer and winter values of this fluctuation, 
the control of the southern winter on the succeeding summer being 
expressed by a coefficient of 0-82, the corresponding data being plotted 
together in Fig. 8; but the relationship with the previous summer is 
