A rationale is presented for the "control items" technique, the technique of using control (nonexperimental) test items to provide a baseline for estimating the effects indicated by scores on the experimental test items. For evaluating the effects of a curriculum, a test can be made of two types of items--ones that test explicit program objectives (program items), and ones that test explicit objectives beyond those of the program and that are also relatively insensitive to transfer (control items). Control items are thus a conceptual analogue of control groups and must be comparable in several respects to program items, just as experimental and control groups must be comparable. Precedents for control items are given, and potential advantages and disadvantages of control items are discussed. They can be used alone or in conjunction with control groups or time series. Sources of possible control objectives are discussed. A number of issues still to be resolved are mentioned, and questions are asked which in this context may suggest new solutions to some enduring measurement and design problems. (Author/CTM)