Mastery tests are analyzed in terms of the number of skills to be mastered and the number of items per skill, in order that correct decisions of mastery or nonmastery will be made to a desired degree of probability. It is assumed that a random sample of skills will be selected for measurement, that each skill will be measured by the same number of items, and that there is a predetermined passing score. There are also implications about the number and quality of the distractors that are used. In this analysis the goal is to determine whether the proportion of skills that an examinee knows is above or below a known constant. The goal is to find a conservative solution to the choice for the number of skills to be sampled so that the probability of a correct decision is reasonably close to one, regardless of the examinee's actual ability level. Illustrations are given to demonstrate that the consideration of errors at the item level might have a substantial effect on the number of items needed. Another approach to the same problem, which is based on latent structure models, is outlined. (Author/CTM)