468 
NATURE 
[APRIL 7, 1923 

even two or three at the time; there may be half-a- 
dozen variables. 
plicates any attempt to apply to field conditions the 
results obtained by these single factor physiological 
experiments. It is possible that when the physio- 
logists have completely elucidated all the single factors, 
some one will be able to synthesise the material and 
build it into some great conception or expression that 
will contain all, and thus account for the field results. 
But history shows that the genius capable of effecting 
a synthesis of this sort is very rare and might have 
to be awaited long. 
We. have therefore adopted another method at 
Rothamsted, which is being worked alongside of the 
single factor method. Statisticians have, during recent 
years, been evolving methods for dealing with cases 
where several factors vary simultaneously. These 
methods have been applied by Mr. R. A. Fisher to the 
Rothamsted field data, and he has been able to trace 
KZA Siow Changes 
E2esi,,. soit Deterioration 
MR Annual Changes.—Weather 
AB= mart predictable from ram record only, 

3 i) 
NO MANURE FARMYARD COMPLETE NO POTASH 
MANURE ARTIFICIALS. 
Fic. 2.—Mr. R. A. Fisher's results showing amounts of predictable 
variation in wheat yields, Broadbalk, Rothamsted. 
certain statistical regularities which foreshadow the 
possibility of important developments. 
Thus, the yields on the Broadbalk wheat field vary 
every year, apparently in a most erratic manner. 
But analysis of the figures showed that the factors 
causing variation could be disentangled and expressed 
quantitatively ; there are slow changes in the field, 
such as changes in the amount of weeds, etc. ; deteriora- 
tion of soil; and weather changes such as rainfall, 
temperature, etc. (Fig. 2). 
As might be expected, the effects differ according 
to manurial conditions ; e.g. the influence of weather 
varies with the manure. Important differences appear 
between farmyard manure and artificials. The varia- 
tion in yield is less where farmyard manure is given 
than where artificials are used. Further, the so- 
called complete manure appears not really complete 
at all; there is soil deterioration going on; but with 
farmyard manure no such deterioration is produced. 
The different kinds and quantities of artificial manures 
produce different effects on the variation in yield, 
the magnitude of which has been worked out. 
Having disentangled the factors Mr. Fisher has 
NO, 2788, VoL. 111 | 
This, of course, enormously com- | 

proceeded to analyse the effect of rainfall, and he 
finds that part of the weather effect is predictable 
when rainfall is known. Rain above the average in 
autumn is somewhat beneficial; in winter and in 
summer it is harmful, and in spring it is less fre- 
quently harmful. As before, the effects are much 
more pronounced with complete ar‘ificials than with 
farmyard manure. The actual facts have long been 
known in a general way, but here is an exact quantita- 
tive measurement. 
The great advantage of this statistical regularity is 
that it indicates the possibility of expressing in terms 
of chance the influence of the weather, soil, etc., on 
crop yields. We hope ultimately to be able to say 
to the farmer, given such and such conditions of soil 
and weather, the chances are so many to one that such 
and such an increase of yield will be obtained by the 
use of a specified fertiliser. The expression would 
be understood by every farmer, and he would readily 
decide whether to take the risk or not. 
Much greater results would also follow. At present 
the farmer cannot cover his risks of low yields by 
insurance, the companies not yet having sufficient 
data. We hope and believe that these statistical 
investigations will afford the basis on which such 
data will be obtained. At present the position ap- 
proximates to that of life insurance in the eighteenth 
century, when the statistical regularity of mortality 
was first established, after which the first life tables 
could be constructed. There still remains a-mass of 
detail to work out, but the fundamental problems are 
now being attacked, and we see no reason to regard 
them as insoluble. If the expectancy of crop yields 
proves to be calculable the farmer will be able to 
insure himself against crop failure, and so meet one 
of the worst vicissitudes of his troubled life by merely 
taking out an insurance policy—perhaps even by 
subscribing to a particular newspaper. 
We are constrained to admit that the work is still 
far from completion, and in the meantime agriculture 
has fallen on difficult days and farmers are turning 
to us to ask how they can obtain large crops in the 
most economical way. It is not general principles 
they want, but particular instructions. 
We are not in a position to give an absolute clear- 
cut prescription to any farmer, but we are going a 
long way to meet him. Some of our field experiments 
of special interest or importance are being repeated 
at other centres where soil and climatic conditions 
are all different. The results are compared with ours, 
or with others that have been obtained, to ascertain 
how far or in what direction any of our conclusions 
would need modification in a particular district. 
We now return to an important result to which 
I have already referred. Over a period of years the 
artificial manures have not proved quite as effective 
as farmyard manure; there has been more variation 
in yield and they have not so well maintained the 
fertility of the soil as tarmyard manure has done. 
On some crops the effect is marked ; clover responds 
better to farmyard manure than to artificials. It 
appears then that Lawes’ and Gilbert’s views that 
the fertiliser value of farmyard manure lay in its 
ash constituents plus nitrogen compounds is only a 
first approximation, and that farmyard manure does 

: 
