PREFACE TO THE FOURTH EDITION vii 



dealing with a single substance can treat it in a certain 

 way, and be reasonably certain that the result obtained 

 is the direct consequence of the treatment. The soil 

 investigator has no such certainty : definite treatment 

 of a soil may be followed by a definite result, but there 

 may be no direct relationship with the factor under in- 

 vestigation the result may be merely a reflex of some 

 far-reaching change produced in some other factor which 

 is entirely overlooked. Further, if observations are 

 attempted in the field it is impossible to ensure such 

 simple variation as the recorded data seem to suggest. 

 It follows that the ordinary laboratory method in which 

 factors are varied only one at the time requires con- 

 siderable modification when used for soil work. 



The methods in use at Rothamsted fall into two 

 groups : 



1. Observations are made in natural conditions as ac- 

 curately as is feasible, and repeated sufficiently frequently 

 to allow of treatment by modern statistical methods. 

 These enable the investigator to study the variations, 

 and hence to make deductions as to the numbers and 

 properties of the factors involved. The factors can then 

 be studied in the laboratory as single factors, using more 

 precise methods and more rigid controls than are possible 

 in the field. 



2. Experiments are made on the soil, and from the 

 results deductions are drawn as to' the probable nature 

 of some new factor. Direct experiment is then made 

 to test the operation of the factor in the field, and precise 

 laboratory experiments are also undertaken. 



Further, just as the ordinary methods of investigation 

 are insufficient, so also the customary divisions of science 

 cannot be rigidly maintained in soil work. The chemist 

 is constantly confronted with physical and biological 



