With the variety Snyder a difference of 2.4 grams in weight per nut in samples 1 to 5 suggests poor sampling technique as this is an objective value. A difference of 4.2 per cent in first crack suggests carelessness on the part of the operator in cracking or difference in soaking as this is quite out of line with the variation of .8 per cent in per cent weight of total kernel. The difference of 16 quarters is considerable but represents only 1.6 score points. As with the Spear the variation in penalty of 4 points is greater than other factors except per cent first crack (i.e. 4.2% points). The difference in score of 5.5 points is obviously greater than desirable, but probably indicates the relative value of the samples. Without penalties the difference is 4.5 points. Sample 7 of Spear and number 6 of Snyder were cracked November 4th when only partly cured and show the importance of curing in obtaining an accurate rating for a sample. The score of each variety was increased materially in all characteristics and no shrivelling was apparent. As a practical means of recovering the kernels in large pieces, cracking before the nuts are dried out is a decided advantage provided the kernels are cured before they are stored. The duplicate samples of Eldridge check very closely and show no significant differences. In Table 2 are given the results of ten tests on carefully replicated random samples of Snyder black walnuts. In making these samples the nuts were spread in a single layer on the floor and lots of 25 cut off the edges of this layer without selection of any kind. Even with such selection there is a variation of 1.2 grams in the average weight of single nuts from different samples. Per cent kernel first crack shows a minimum of 21.8 and a maximum of 26.9 in the ten samples. This difference is related mostly to the presence of 3 empty nuts in the low scoring sample as compared with none in the high scoring sample. The high score is also in part due to soaking. This variability is about the same as with total per cent kernel indicating that cracking technique was uniform. Comparing samples 1 and 2 in more detail it is found that the difference of 11.6 points in the score is caused by the presence of empty nuts in the sample. The average weight of kernels per single nut in sample 1 is 4.9 grams. The difference in the weights of the kernels of the two samples is 15 grams or about the weight of the kernels of 3 nuts. These empties also reduce the score by reducing the number of quarters recovered. Where empty nuts are involved, it is doubtful if random sampling will reduce variation unless the size of the sample is greatly increased, a practice which is not a practical solution in that a 25 nut sample is about as large as can be handled with any facility. It would seem that this difference in scores was a fair indication of the merit of the two samples. The scores of the other samples show a fair degree of uniformity.