278 



Journal of Agricultural Research 



Vol. Xn, No. s 



The coefficient of variability for the lo-tree plot is remarkably low 

 for the reason that the trees themselves are so uniform and only a small 

 area of ground is involved. It is no surprise, therefore, to find that the 

 coefficients of variabiUty are nearly equal, calculated from one to six 

 years. The probable errors are relatively large and it is difficult to 

 assert that there is any real difference. 



The walnut yields can be used as additional data for the study of this 

 question. Table XVI shows the coefficient of variabihty of the 191 5 

 and 1 91 6 yields and their total, with both i- and 8-tree units. By con- 

 sidering the individual tree as a unit, the total yield for the two years 

 was less variable than the 191 6 yield, the difference equaling three times 

 the probable error. The variability of the 191 5 yield is practically 

 equal to that of the total. On considering an 8-tree plot as a unit, there 

 is a difference between the coefficients for 191 5, 191 6, and the total of 

 the two years respectively ; however, the observed difference is less than 

 three times the probable error and its significance may be somewhat 

 doubted. Apparently the mean of two years' yields in this case is less 

 variable than one year's yield. 



Table XVI. — Comparative variability of yields of seedling walnut trees through a 



period of two years 



Class. 



Individual trees 



Plots of eight trees each. 



Coefficient of variability of the yield. 



191S yield. 



47- 9±i-6 

 30. 0±2. 6 



1916 yield. 



53-9±i-9 



Mean of 



191S and 1916 



yields. 



46. 4±i. 6 

 25. 3±2. 2 



Further studies on the comparison of the variability of yields through 

 several years were made, from data published by Hedrick (igii). Table 

 IV of the bulletin cited gives the yearly yields of individual apple trees 

 from 1902 to 1 910, inclusive, upon which our computations are based. 



Four differential fertilizer treatments have been given to eight plots 

 of five trees each, each treatment being duplicated on nonadjacent plots. 

 There are three nontreated plots serving as controls. 



The variability of the individual trees was computed on the 1 5 control 

 trees. 



