No. 583] ON SUBSTRATUM HOMOGENEITY 445 



tion plots each of 2 X 4 = 8 ultimate plots. The method 

 of dividing up the field is indicated by the marginal ar- 

 rows on Map C. 



S[n(n - 1)] = (12 X 16 X 15) + (4 X 8 X 7) = 3104. 

 For grain, 



S[(n-l)p] =1707635, #[(»- l)pP} =9683408.57 

 p = 550.140, <r p 2 = 9311.307, 

 S{C P 2 ) =1023184887, S(p 2 ) = 70112319, 



whence 



r PlP2 = .472 ± .035. 



For nitrogen, 



8[(n-l)p-] =6458.63, S[(n - l)p 2 ] = 13464.6031, 

 p = 2.080744, * P 2 = .008327, 

 S(C P 2 ) =14409.6095, S (p 2 ) =968.3721, 



and 



r PlP2 = .096 ± .045. 



Again the weighted means and standard deviations do 

 not differ widely from those used above. The differ- 

 ences in the correlations will be discussed below. 



In concluding this section it may be pointed out that 

 all of the foregoing values are surprisingly high. They 

 indicate clearly that the irregularities of an apparently 

 uniform field may influence profoundly the yield of a 

 series of experimental plots. They also bring out an- 

 other interesting point. In the three cases in which two 

 different characters were measured on the same species 

 they show very different susceptibilities to environmental 

 influence. Thus, for example, the correlation of man- 

 gold roots is r = .346 + .042 as compared with r = .466 + 

 .037 for leaves. For grain on the Rothamsted field with a 

 4 X 5-fold grouping the correlation is r = . 186 + .029 as 

 compared with r = .343 + .027 for straw. For Montgom- 

 ery's data for yield and composition the differences are 



