The objectives of this investigation are to develop, describe, assess, and demonstrate procedures for constructing mastery tests to minimize errors of classification and to maximize decision reliability. The guidelines are based on conditions where item exchangeability is a reasonable assumption and the test constructor can control the number of items on the test and its cutoff score. Eight steps for minimizing errors of classification are described. The major difficulty of the procedure is the identification of an appropriate criterion variable. Suggested is the identification of a reasonable concurrent criterion variable in order to set temporary standards of mastery until sufficient longitudinal data can be obtained. The procedure for maximizing decision reliability involves five steps. A shortcut to the procedure is provided in the form of tables for determining the proportion correct for mastery for tests ranging in length from five to ten items. A major advantage of the procedures described for generating mastery test forms from an item bank is that they are well within the reach of most local school districts desiring accuracy in mastery assessment programs. The procedures do not require unduly complicated computational procedures. (Author/GSK)