ABSTRACT
A test is rarely given with a single purpose in mind. For example, aside from their intended use, scores on college admissions tests like the SAT and the ACT are also used in related tasks like awarding scholarships, as well as for completely unrelated tasks like the evaluation of states' educa tional policies. Sometimes multiple uses of the scores are complementary, and the same test construction procedures that build a test that is good for one purpose also yield a test that works well for the others. Sometimes, however, the goals are antithetical: A test that is built to precisely rank examinees is likely to be rather poor at providing specific diagnostic infor mation for remediation. An ability test that is built to measure an ex aminee's potential for learning may generalize far more broadly than a performance assessment that is built to measure how well an examinee can
perform a specific task. Conversely, if we want to know if someone can change a tire, it is best to ask that they do exactly that, rather than use the indirect and inefficient approach of administering a general test of me chanical aptitude.
