452 



tuk ami- in t i.\ \.\ iri: \i.i>r [Vor,XLix 



V. Recapitulation and Discussion 



If the methodical production of new varieties of animals 

 and plants to be made possible by the laws discovered 

 in experimental breeding is to be of material practical 

 value, more attention must be given to the development 

 of a standardized scientific system of variety testing. 

 From the practical standpoint, nothing is to be gained by 

 the formation of varieties of plants differing in discern- 

 ible features of any kind unless some of these varieties 

 can by rigorous scientific tests be shown to be of superior 

 economic value. 



It is equally true that if tests of fertilizers or of dif- 

 ferent methods of irrigation carried out on an experi- 

 mental scale are to have any real value as a guide to a com- 

 mercial practise, the differences in the experimental re- 

 sults must certainly be significant in comparison with 

 their probable errors. 



The problem of plot tests has several different phases, 

 all of which must ultimately receive careful investigation. 

 The purpose of this paper has been to consider one of the 

 problems only. To what extent do the irregularities of 

 an apparently homogeneous field selected for comparative 

 plot tests influence the yield of the plots! 



The question has been far too generally neglected, 

 although indispensable to trustworthy results. It is ob- 

 viously idle to conclude from a given experiment that va- 

 riety A yields higher than variety B, or that fertilizer X 

 is more effective than fertilizer Y, unless the differences 

 found are greater than those which might be expected 

 from differences in the productive capacity of the plots 

 of soils upon which they were grown. 



The first problem has been to secure some suitable 

 mathematical criterion of substratum homogeneity (or 

 heterogeneity). Such a criterion should be expressed on 



