FRUIT BREEDING 



267 



From these results the conclusion is reached that eight trees is 

 about the correct number which should be used in a plot. 



The question of replication, i.e., the systematic distribution of 

 plots over the field, is taken up. Results computed for four- and 

 eight-tree units are given for oranges, apples, walnuts, and 

 lemons. Table LXX gives an average of data from these crops. 



TABLE LXX. EFFECT OF REPLICATION IN FOUR- AND EIGHT-PLOT UNITS 



The conclusion seems warranted that four systematically 

 replicated plots greatly reduces the error which arises from soil 

 heterogeneity. The data also show that four systematically 

 distributed plots of four trees each are somewhat more reliable 

 than two plots of eight trees each. 



As was presented in Chapter IV, Harris has given a reliable 

 means of estimating soil heterogeneity by the correlation between 

 the neighboring plots of a field. The test was applied to an 

 orange grove which' appeared to have uniform soil conditions. 

 The correlation between the yield of eight-tree plots as ultimate 

 units and grouped combinations of four such adjacent plots 

 was found to be: r= +0.533 + 0.085. This showed a pro- 

 nounced heterogeneity in the soil of this orchard. However, the 

 correlation computed between the yield of an eight-tree ultimate 

 unit and the yield of the combination of four such systematically 

 distributed units was not much larger than the probable error. 



These facts show the unreliability of yields of single trees as a 

 criterion of productivity, that eight-tree plots give much more 

 reliable results, and that plot replication is of as much value in 

 studies of fruit-yield as of farm crops. Where quality is a major 

 criterion, single trees give fairly reliable information. 



Self -sterility and Heterozygosity. One of the chief difficul- 

 ties of systematizing methods of work is due to the heterozyogus 



