A number of factors can influence the decision on how many items to sample. Of major importance is the level of confidence with which the recovered items can be regarded as having originated from the same source as each other, prior to performing the particular type of comparison with a control. The binomial distribution is appropriate for considerations for a very large total number of items studied, but when a finite number of items is involved, the hypergeometric distribution should be used. The probabilities of randomly selecting certain numbers of matching items in sample sizes of 5, 8, and 10 from total numbers of recovered items made up of different proportions of matching and non-matching items are shown. These numbers are in the range that allows for the tables to be applied to many of the situations that arise in case work in which the comparison of glass is involved. Relatively simple calculations carried out using this model distribution can help illuminate the value and the limitation of sampling. 1 table and 1 reference