100 BULLETIN 624, U. S. DEPARTMENT OF AGRICULTURE. 
In each succeeding year of these investigations, the workers be- 
came familiar with an increasing number of variable forms of fruits 
and recorded their occurrence as they were observed. This accounts 
for the general increase from year to year in the number of such 
fruits recorded. Because of this increase in the number of variable 
forms observed, the figures given here as the average annual pro- 
duction of such fruits are too low, especially for those trees which 
have borne a large number of such variable fruits. When securing 
performance records, late-bloom fruit and split fruit have been re- 
corded, but such fruits have not been included with the variable 
fruits. 
In expressing the average of weights in Table I, it was found 
impracticable to retain more than one decimal place. The more 
exact expression of ounces as a fractional part of a pound extends 
to four decimal places, but only one place has been retained. In 
expressing the averages for the number of fruits occurring in differ- 
ent groups, no decimal has been retained except when the average 
number is less than unity. Hence, it will be found that the totals of 
averages sometimes will vary slightly from the average of the totals 
of the corresponding number. 
In recording the tree-performance data the fruits of the Cull grade 
are not assorted into sizes, and on this account the total figures for 
the weights and numbers of fruits of the various sizes represent only 
the commercial crops of the trees. 
In determining the average number of seeds per fruit on the differ- 
ent trees counts are made when possible from a small, a medium, 
and a large fruit in each of the three grades. In this way a record is 
usually secured of the actual number of seeds in 9 fruits from each 
tree, and it is thought that this system gives a fairly accurate indica- 
tion of the relative seediness of the fruits of the different trees. To 
determine the annual average number of seeds per fruit, the total 
number of seeds found is divided by the total number of fruits 
examined during the entire period. In the case of trees from which 
less than 9 fruits have been examined during any season the average 
obtained in this way gives equal weight to each fruit and is fairer and 
more accurate than to average the yearly averages. 
In counting the variable fruits occurring on individual trees, the 
strain of each tree is considered the standard for that tree, and all 
deviations from it are considered as variable fruits. For example, 
on a tree of the Valencia strain all fruits other than those of the Va- 
lencia strain are considered to be variable for that tree, while on a tree 
of the Corrugated strain all fruits other than those of the Corrugated 
strain' are counted as variable. With the trees of the Sporting strain, 
however, the fruit of the Valencia strain is considered to be normal, 
and all variable fruits are counted as off type for those trees. The 
