48 LIMITS OF NATUEAL VARIATION. 



Some of these causes of discordance deserve a more special 

 consideration. 



7. Importance of ascertaining the limits of natural variation 

 in the productiveness of an experimental field. Necessity of 

 double experiments. 



No adequate or satisfactory consideration has yet been given 

 to the question When are results to be considered as identical ? 

 I have elsewhere * drawn attention to the importance of this 

 question, and endeavoured to show how the neglect of it alone 

 throws doubt on most of the experimental results hitherto pub- 

 lished. A farmer divides his field into four parts : to one he 

 applies nothing, but to each of the three others he applies a cer- 

 tain quantity of three different substances. From No. 1 he 

 reaps a certain amount of crop, and from each of the others an 

 increase greater or less according to the substance applied. It 

 has been usual to ascribe the differences between the crops on 

 each of these latter and the crop of No. 1 to the effect of 

 the substance applied to measure the several effects of these 

 substances by the several differences observed. Is such a mode 

 of procedure correct? Are the inferences or measurements 

 likely to be just ? Experience, I think, gives a negative answer 

 to these questions. Thus 



1. In 1843, Mr Dockar, of Findon Farm, Aberdeenshire, made 

 experiments upon turnips (green-topped yellow) with and with- 

 out manure. From two separate eighths of an acre of the same 

 field, to which no manure had been applied, he obtained at the 

 rate of (tops and tails excluded) f 



First portion, . . 8 tons 11 cwt. per imperial acre. 

 Second portion, . . 6 ... 16 ... 



Difference, . . . 1 ton 15 cwt., or If tons on a crop 



of 6f to 8j tons. 



2. At Barrochan, in Kenfrewshire, from two unmanured 



* Journal of ike Royal Agricultural Society, IX. p. 200. 

 t Transactions of the Highland Society, July 1845, p. 7. 



