The feasibility of creating an item bank from a teacher-made test was examined in two comparable sections of a graduate-level introductory measurement course. The 67-item midterm examination contained multiple-choice and master matching items, which required higher level cognitive processes such as application and analysis. The feasibility of banking the test items was determined by the comparability of invariant item difficulty (item reliability) and item quality (item validity). The test was administered to Spring and Fall classes, and a Rasch latent trait analysis was performed on each test administration. The reliability of each item difficulty was determined by comparing the difference between Rasch difficulty estimates from each class for each item against their calibration error. Item validity was studied by examining the extent to which each item approximated an a priori rating used for criterion-referenced grading. Test reliability and test validity results indicated that the majority of items could be usefully banked. In the future, new test items could be created using the same procedures. The development of computer software to facilitate item analysis was recommended. (GDC)