Two basic requirements for the successful initiation of a program for test assembly are the development of detailed item content classification systems and the delineation of the professional judgements made in building a test from a pool of items to detailed content, ability, and statistical specifications in terms precise enough to be translated into computer programs. A guiding principle has been that the computer is to serve the professional staff, not supersede them. Every test assembled is subject to professional review; revisions can be requested; and the items selected are held "in limbo" until the test has been accepted. The systems development turned out to be much more sophisticated than originally expected. It is undoubtedly more complex than would be required for most testing situations since it is designed to handle a wide variety of tests in each subject area and the numerous constraints that are imposed on item selection by the rigorous test specifications and need for parallelism of forms that is basic to some of Educational Testing Service's (ETS) national programs like the College Board tests. In addition to a unique identification number and the classification for each item, the computerized file contains a complete history of the item's uses, up to five sets of statistics, codes for the security level and present activity status of the item, and up to 15 12-letter key words--which in the case of some verbal items constitute the complete items. (Author/BJG)