354 Journal of Agricultural Research voi.xx.No. s 



insignificant negative correlation observed may be due to some pecu- 

 liarity of the crop. The comparison of the correlation for the 191 5 and 

 1 91 6 corn crops has already been made (Table XI). Inspection of the 

 averages in Table XII shows that on whatever character they are based 

 the correlations decrease from the maximum relationship observed in 

 1 91 6 to the lowest values in 191 9. # 



Whether the residual influence of alfalfa per se has any influence on 

 the 1 91 9 or later crops can only be determined by further experimenta- 

 tion in which the interannual correlations can be deduced from the 

 yields of plots upon which alfalfa has not been grown. 



III.— DISCUSSION AND RECAPITULATION 



The purpose of this paper has been to present the results of a new 

 method of attack upon the problems of (a) the permanency of the differ- 

 ences which are found in the plots of an experimental field, and of (b) 

 the influence of variations in the yields of certain crops in the rotation 

 upon the yields of subsequent crops. 



The data upon which the studies were primarily based comprise the 

 yields of 46 plots — subdivided in several cases into half plots and quarter 

 plots — each of 0.17 acre in area at the Huntley (Mont.) Field Station of 

 the Office of Western Irrigation Agriculture for the nine years between 

 191 1 and 1 91 9, inclusive. 



The uniform cropping experiment, involving sugar beets, alfalfa, 

 corn, oats, and barley, was initiated merely to determine the variation 

 in the yields of plots of a given size when homogeneously planted and 

 uniformly treated. The experimental procedure was, therefore, deter- 

 mined in advance and was wholly independent of the statistical analysis. 

 This is in certain regards fortunate. It frees the data absolutely fronf 

 any suspicion of an influence of preconceptions or of personal equation 

 on the biometric results. On the other hand, it is quite possible after 

 the statistical analyses have been made to recognize ways in which the 

 experiments could have been improved and made to yield more valuable 

 results. This is, however, a feature of research in general. The dis- 

 covery of inadequacies in a first set of experiments makes possible their 

 elimination in subsequent work. The most unfortunate defect in the 

 data was that the harvesting and weighing could not be done by half 

 and quarter plots in 1917, 1918, and 1919, but this curtailment could 

 not be avoided under existing conditions. 



The results of a previous study (3) have shown that fields selected for 

 plot tests of all kinds are practically without exception heterogeneous 

 to a degree that influences profoundly the yields of the crops grown 

 upon them. It was there pointed out that the correlation between 

 the yields of adjacent plots might either be due to initial physical and 

 chemical differences in the soil or be referable to the influence of previ- 

 ous crops upon the composition, texture, or tilth of the soil-. 



