Feb.4,i9i8 Variability of Yields of Fruit Trees and Field Trials 271 



used, but when controls are placed every second or third plot in the 

 series, the evidence is in favor of the long and narrow plots. The use of 

 every second or third plot in the series as a control plot obtains greater 

 accuracy than when no controls are used and the average of the field is 

 considered the normal yield for all plots. 



A long plot is much more economical of time and labor than a square 

 plot containing the same number of trees. If the orchard must be irri- 

 gated, a square plot containing nine trees will require three standpipes 

 instead of one and also three times as many irrigation furrows. 



Arguments have been advanced from time to time in favor of both 

 the linear and the square plot. The advocates of the former have 

 urged its reliability on two points: first, if the soil or other conditions 

 change in a direction parallel to the plot, it will contain both high- and 

 low-yielding areas, and the average will correct one or the other error; 

 second, if conditions change in a direction more or less perpendicular to 

 the plots, each plot will vary from its adjoining plot because of its shape 

 by such a minimum that intercomparisons are more reliable. On the 

 other hand, the advocates of the square plot have claimed that the 

 arguments apply equally well to their case, provided the plot be made 

 small enough. 



In view of the importance of the shape of the plot for cultural opera- 

 tions, the writers have investigated the variability of plots of various 

 sizes and shapes in three of the experimental orchards. The results 

 (Table X) are for unrepeated plots and for one year's yield of fruit, 

 except the apples, which are for two years. Considerable interest lies 

 in the computation on the 9-tree plot in the total navel-orange grove, 

 because it is based on a large population (i,cxx) trees) and because the 

 soil is known to vary in a general way from northeast to southwest. 



Comparisons were made between square plots containing 9 trees and 

 linear plots consisting of a single row of 9 trees extending in the 

 north-south and in the east-west direction. It is interesting to see how 

 closely the coefficients of variability coincide in the three cases. The 

 differences between the coefficients are only a fraction of the probable 

 errors. Therefore on this lot of i ,000 orange trees there is no difference 

 between a square and a linear plot of 9 trees, so far as the reliability 

 of comparative yields is concerned. (The deviations were taken from 

 the mean of all the plots of the grove.) 



Similar comparisons were made on a selected block of 256 trees in this 

 same grove. The coefficient of variability of the trees taken singly is 

 34.47 ±1.14, which indicates that these trees were a fairly typical sample 

 of the entire grove. The block was divided into plots of various shapes 

 containing 16 trees. With one exception there is little difference in the 

 variability of the plots, regardless of their shape. This exception is 

 found in the linear plot of i by 16 trees extending from east to west, 

 which has a lower coefficient than any other arrangement. In the block 



