fect the score.

The presence of empty or poorly filled nuts in a lot of nuts from which
samples are taken at random introduces greater variability in the
samples than that found in lots with all nuts filled. This is true
because the chances of getting an equal number of empty nuts in 25 nut
samples are small and the presence of each empty nut decreases the per
cent kernel and also the numbers of quarters possible. Variations due to
empty nuts could be eliminated by greatly increasing the number of nuts
in the sample but this is not practical for the purposes this schedule
is intended to serve.

The question of whether or not it is possible for different operators to
obtain equal scores on duplicate samples is not satisfactorily answered
by the data in table 4. As the data stand the scores are far from equal.
There is, however, a consistency in the scoring of each operator and it
is quite probable that with more uniform treatment of nuts before
cracking and more careful sampling better agreement would be achieved.
This is borne out in the data given in table 5 in which the variation in
scores between the two operators was no greater than that obtained by
the same operator.

From a study of the data secured it appears that the causes of variation
in the scores of duplicate or replicate samples are the result of (1)
lack of care in making replicate random samples, (2) differences in
treatment of samples before cracking, particularly as regards moisture
content, (3) differences in the skill or care of the operator making the
tests, (4) presence of empty nuts or shrivelled kernels in the sample
which introduces variation not compensated for in a 25 nut sample and
further complicates the matter because assigning penalties for
shrivelled kernels involves personal judgment.

The first three of these can be minimized or eliminated by care and
skill. The fourth item is not so easy but procedure can at least be
standardized. Increasing the size of the sample is not practical if much
testing is to be done.

All things considered it would seem that the scores indicate fairly well
but not accurately the relative merit of the samples and thus can be
relied upon to determine the relative merit of a variety or clone, the
suitability of the variety for growing in a given locality and the
variability of a variety grown in the same region but under different
conditions. To determine the merit of a variety as compared to another
both