In considering the value of any schedule the following questions are pertinent:
1. Is it possible for one operator testing one lot of nuts to obtain the same score with replicate random samples?
2. Is it possible for different operators to obtain approximately the same score on replicate samples?
3. Does the score give an accurate evaluation of the variation of a variety from year to year in one locality or in the same year in different localities? The latter is very important in determining the regions to which the variety is best adapted and the performance of the variety in any one locality.
4. What are the causes of variation in the scores obtained? Which of these reflect the inherent worth of the sample and which are related to technique, personal equation and methods of handling the sample?
5. What changes may be made in the schedule to weight the various factors to give a more realistic score of what changes in procedure will make the schedule more realistic?
Table 1 gives data on replicate samples tested by the same operator. In the samples of Spear, numbers 1-6 the variation is as follows: weight of single nut 1.3 grams, per cent kernel first crack 2.9, total per cent kernel 2.6, number of quarters 3, penalties 4.5 points, score 9.2 points. In scores figured without penalty the variation is 5.4 points. Sample No. 7 was cracked November 4 before the nuts were dry and hence is not comparable with others.
Analysis of these differences indicates that the variation in nut weight is closely related to the number of shrunken and empty nuts in the sample. This is a difficult factor to evaluate in a practical way. At the time of the 1939 report it was suggested that the score should be figured on the basis of filled nuts. This cannot be arranged easily in testing because if the operator cracks the nuts before weighing there is almost sure to be loss of fragments of shell. Trying to correct the original weight in any way is necessarily inaccurate. Deciding whether or not the kernel of a nut is sufficiently shrivelled to deserve a penalty is a matter of judgment which is a personal matter.
The variation in per cent kernel first crack and total per cent kernel probably represents fairly the difference in the samples. The total per cent is a wholly objective value and varies practically as much as the per cent first crack. Uniformity in the number of quarters is striking. This large number is undoubtedly related to the fact that many of the kernels were shrunken enough to be penalized and others were perhaps shrunken enough so that they did not tightly fill the shell cavity. In general it may be said that the more tightly the kernels fill the shell the more difficult it is to extract large pieces. Thus having the kernels a little shrunken but not enough to seriously reduce their weight favors a higher score. Of course, in some varieties the kernels may he plump and still not fill the shell tight enough to make cracking difficult. This is a desirable condition.
Variability in penalties is more important (i. e. 4.5 points) than any other factor in influencing the final score. Without the penalties the scores of samples 1 to 6 would be 87.5, 84.0, 83.6, 83.7, 82.1 and 82.8 respectively which is fairly uniform. Statistically the presence of empty or shrivelled nuts in a lot from which samples are taken increases the number required to make a satisfactory sample by greatly increasing the individual variation of the single nut.