For individualized or computer assisted instruction, norm referenced testing is inadequate to determine each individual's mastery on specific kinds of tasks. Hively's item forms and Ferguson's stratified item forms, both based on observable characteristics of the problems, and Scandura's algorithmic technology, positing that persons use rules to solve problems and thus that problems should be partitioned on the basis of rules needed to solve them, have been developed to measure individual mastery. This study was designed to compare their effectiveness and efficiency in assessing mastery of column subtraction problems. All three methods were essentially equal in predicting mastery of individual items, but the algorithmic method used far fewer items and thus was more efficient. The item forms technology would seem to have a slight advantage in the ease with which a computer could randomly generate test items, but even items for the algorithmic form can be computer generated, although slightly indirectly. (RH)