86 WHEAT PRODUCTION IN NEW ZEALAND 



owing to some adventitious aid. Thus the best heads are 

 picked out of the mixed population in a single year, and 

 nothing remains but to multiply the progeny of that 

 head until enough seed is procured for commercial 

 purposes. 



In practice the process is not quite so simple as the 

 above description would suggest, but the lack of sim- 

 plicity is due to that bane of all agricultural trials the 

 experimental error. 



4. Experimental Error.* 



The "Probable Error" and the "Theory of Chances" 

 are branches of mathematics with a very practical bear- 

 ing. Unless it is constantly recognised that there is a 

 probable error, all experimental work is reduced in value. 

 In agricultural experiments the probable error is very 

 large and very elusive. If two plots of wheat are grown 

 side by side in a paddock the soil of which is apparently 

 very even ; if the seed of the two plots is out of the same 

 bag; if the cultivation, manuring, and harvesting are 

 the same in all respects, it might be anticipated that the 

 yields would be the same within a very slight margin. As 

 a matter of fact the yields often differ by quantities 

 that are of the greatest importance. For instance, the 

 land occupied by one plot may be three inches lower 

 than the land occupied by its neighbour. Such a 

 difference could never be detected, but if heavy rain 

 comes while the seed is germinating the seed on the lower 

 plot may be drowned out, or so checked in its develop- 

 ment as to yield much less than the plot on higher ground. 

 Or a colder piece of soil may delay ripening and render 

 its crop liable to rust, or the warmer soil may ripen its 

 crop earlier so that it is attacked by birds, and the final 



*See supplement No. 7 to "Journal of Board of Agricul- 

 ture." London, 1911. 



