E. J. Russell 327 



an attempt is being made to attack the problem on the following lines. 

 The soil is being studied from the points of view of the chemist, the 

 physicist, and the biologist ; under the last heading separate investigators 

 have charge of bacteria, protozoa, fungi, algae, and helminths. The 

 climatic conditions are difficult to study, but a start has been made by 

 following up the work of Hooker at the Meteorological Office, who has 

 indicated certain correlations between weather and crop yields and shown 

 that the subject is suscej)tible of investigation. As regards methods, 

 one of the chief necessities is the collection and interpretation of data. 

 Here one is met by the special difficulty that there can be no clear cut 

 start. A chemist in securing data can at any time begin anew with 

 perfectly clean apparatus and recrystallised initial substances, whereas 

 in agriculture there is always a previous history to take into account: 

 the composition of the seed is affected by the state of the parent plant, 

 and soil may show the result of treatment after the lapse of many years. 

 One of the plots at Rothamsted is still affected by manurial dressings 

 which were discontinued in 1870. An experiment therefore is only valid 

 when it is done on fields of known history. 



But this does not end the difficulties. Scientific investigators are 

 brought up on the Baconian method in which all factors are kept constant 

 except the one under investigation. This method has led to wonderful 

 achievements in the laboratory and is insensibly adopted by workers in 

 all branches of science. In field experiments in our uncontrollable 

 climate, however, it cannot be strictly applied. It is true that two plots 

 may be laid out on which all humanly executed processes such as culti- 

 vations and other treatments may be uniform with only one factor, 

 e.g. the supply of a particular fertiliser, deliberately varied. But 

 the phenomena shown by the plant are not necessarily effects of this 

 one factor: they may be caused by a totally difierent factor which 

 has resulted from the interactions of the factor under experiment wnth 

 some set of soil or climatic factors adventitiously coming together perhaps 

 for the first and last time. Fortunately within recent years a method 

 has been evolved for dealing with cases where several factors vary 

 simultaneously. This is the correlation method, but its application 

 involves the use of mathematics beyond the equipment of the average 

 scientific investigator. The method also requires considerable masses 

 of data. 



At Rothamsted the data are all examined by a competent statistician 

 who works out the correlations between the observed effects and the 

 various factors concerned. His results are expressed in curves or in 



