364 EXPERIMENTS WITH FERTILISERS [chap. 



as an error of 10 per cent, or so must be expected in 

 the returns from a single plot, this error must be taken 

 into account in the design of any scheme of field 

 experiments. 



For example, if trial plots are being laid out and are 

 only expected to continue a single year, it would be 

 useless to include among them a comparison of sulphate 

 of ammonia and nitrate of soda containing equal 

 amounts of nitrogen. There is abundant evidence that 

 the superiority of the nitrate of soda is somewhere 

 about 10 per cent. ; but as this is no more than the 

 expected experimental error a single experiment must 

 be inconclusive. If it is important to settle for the 

 particular soil the relative value of nitrate of soda and 

 sulphate of ammonia more plots must be given up to 

 this one question ; at least five would be needed, and 

 even then there would remain a possibly considerable 

 error due to the season. 



This suggests that the prime consideration in 

 designing a set of field experiments should be to limit 

 the scheme strictly to certain definite questions which 

 can be answered in the time and space available. 

 There should be no haphazard laying-out of plots with 

 all sorts of variations of manuring ; the problems to be 

 solved should be clearly thought out beforehand ; it 

 will generally be found that only a very small number 

 of problems can be attacked at one time and every 

 plot should be arranged to contribute to the result 

 without the introduction of any secondary or disturbing 

 factors. 



As to the choice of land for experimental plots little 

 can be done beyond exercising ordinary discretion in 

 selecting a field which promises to be uniform. The 

 geological drift map should be consulted, and places 

 marked by thin patches of drift or on the boundaries of 



