250 Journal of Agricultural Research voi. xii, No. s 



The use of controls and repeated plantings in varietal tests was studied 

 by Pritchard {1916) in breeding work with sugar beets. His studies lead 

 to the conclusion that the practice of dispensing with control rows and 

 using the mean of all progeny rows as a standard of comparison appears 

 to be less accurate than the employment of frequent controls. However, 

 the employment of every alternate row as a control was not sufficient to 

 offset the variability in yield arising from irregularities of soil. 



Stockberger {1916) discussed the value of a number of the common 

 methods for determining the normal yield of treated plots based upon 

 the yields of hops. Normal yields for various plots varied widely accord- 

 ing to the method of computation, the values in some cases differing from 

 the actual yield by as much as 40 per cent. Repetition brought about 

 a very marked reduction in variability, although with only five repetitions 

 the error is still relatively large. 



The work of Surface and Pearl {1916) shows an advance in the refine- 

 ment of methods of conducting field trials. With the realization that 

 the use of frequent control plots often produces results far from satis- 

 factory, these workers have calculated by the contingency method the 

 probable yield of each plot of ground in their grain-testing series. This 

 calculated yield represents the most probable yield of each plot on the 

 supposition that they have all been planted with a h)'pothetical variety 

 whose mean yield is the same as the observed mean of the field. This 

 "calculated" yield may then be used as a basis for determining a cor- 

 rection factor, whereby each area must be given a handicap plus or 

 minus the actual yield, depending upon whether the plot in question is 

 calculated to be a low- or a high-producing area. This method of cor- 

 recting the soil variation is combined with four systematically repeated 

 plots of one-fortieth acre of each variety, and gives a high degree of 



accuracy. 



MATERIAL USED FOR STUDY 



The studies to be reported in this paper deal with the variability of 

 fruit-tree yields. They are based upon the individual tree yields of 

 oranges {Citrus sitiensis), lemons (Citrus limonia) , walnuts (Juglans regia), 

 and apples {Malus sylvestris) from orchards which had received uniform 

 treatments for a number of years — indeed, so far as known, from the 

 time of planting the trees. The orchards were carefully examined, and 

 the records for all trees which were known to be abnormal from disease 

 or other apparent causes were eUminated. In place of the records of 

 trees thus eHminated the average yield of the eight surrounding trees 

 was substituted (assuming that the tree stood at the center of a square 

 block of nine trees). 



This substitution is not entirely satisfactory, yet it was felt that it 

 was necessary in order to compute plots of homologous size and syste- 

 matic arrangement. The writers found, as a matter of fact, that there 

 is a very high degree of correlation in these orchards between the yields 



