72 WHEN EXPERIMENTS SHOULD BE REJECTED. 



trouble, and expense, should, by the chances of an unusual 

 season, by the unsuspected condition of his land, or by other 

 accidents, be rendered wholly abortive. Such accidents form 

 one of those numerous sources of delay to which the progress of 

 scientific agriculture is peculiarly liable, which have made its 

 advance so slow, many of its steps in advance so doubtful and 

 insecure, and have disheartened and driven from its service 

 many useful and talented men.* 



8. When experiments are to be rejected. Value of contra- 

 dictory and of positive and negative results. 



From what has already been said, the reader will be satisfied, 

 I think, that field experiments are not to be confided in, where 

 the limits of natural variation in the land itself have not been 

 previously ascertained by duplicate or triplicate experiments on 

 the soil without addition, or where, this having been done, only 

 one comparative experiment has been made with the substance, 

 the action of which it is intended to investigate. There are 

 other reasons, also, as the reader must have seen, which will 

 lead him at once to conclude, that entire series of experiments 

 are to be rejected as suspicious as unfit, therefore, to be trusted 

 to, to find a place in our books, or to be allowed any weight in 

 our reasonings. This subject is, I think, deserving of special 

 illustration. 



A series of experiments, published or unpublished, ought, as 

 it appears to me, to be rejected 



1. When their results fall within the limits of natural varia- 

 tion. We have seen that two portions of the same field may give 

 a produce of corn, hay, or turnips, under the same circumstances, 

 differing from each other. In the case of 



a Turnips, as much as 25 per cent. Differences which do 

 not exceed 5 or 10 per cent, therefore, may be considered as 

 natural differences ; and without further experiments upon the 

 point, conclusions can scarcely be drawn with safety from com- 

 parative results which approach each other so closely as this 5 

 or 10 per cent. 



* For an instance of such accidents,, see Transactions of the Highland Society 

 for March 1849, -pp. 500, 501. 



