76 CONFLICTING CAUSES MUST BE ELIMINATED. 



e And on the same farm, in the same year, 



24 carts of dung, and 2 cwt. of guano, gave 28 tons 7 cwt. 

 24 3 ... ... 25 ... 4 ... 



Being a difference of .... 3 tons 3 cwt. 



Such sets of experiments as these three last are mere rubbish, 

 and I quote them here as illustrations of the kind of results 

 which ought not to be given to the public. To the individual 

 who made them they may be useful in showing that his methods 

 of experimenting are in error somewhere, in leading him to the 

 cause of his contradictory results, and in preparing him for 

 making trustworthy experiments hereafter ; but the public 

 ought to have been spared the infliction of studying them.* 



The last two experiments made at Kerry tonlia might have 

 been adduced as illustrations of the extent to which farm-yard 

 manure may differ in fertilising virtue a fact to which I have 

 adverted in a previous section. But such an illustration would 

 have been open to the objection, that we have no experimental 

 proof that the different parts of the field might not alone, and 

 without manure, have given returns widely differing from each 

 other. 



We have, in fact, three elements or unknown quantities in- 

 volved in these experiments the soil, the farm manure, and 

 the guano, and only one pair of experimental results from which 

 to deduce the action of each of the three. It is impossible to 

 say how much of the difference observed is due to each. Such 

 experiments, therefore, only uselessly load our books, and give 

 additional labour to those who wish to extract, from the results 

 of practical men, experimental data by which to guide their 

 researches, and test the opinions they may form. I may here 

 remark that, in all experimental inquiries, it is of the greatest 

 possible consequence to eliminate every conflicting cause to 

 remove every circumstance which may interfere with the deter- 

 mination of the precise point which we wish to ascertain. This 

 is very difficult, and, as a general rule, will require at least two 



* See also chapter x. 4. 



