Despite their advantages over other assessment techniques, current achievement and ability tests are not especially efficient sources of information for the range of educational decisions for which they are used and relied upon. Two major types of tests, criterion-referenced and norm-referenced, and two types of use, student evaluation and program evaluation, are considered in this context. The strengths and weaknesses of criterion and norm-referenced measures are discussed in detail. A four-step approach to test construction is proposed, combining the better components of the criterion and norm-referenced approaches, which may overcome some of the information problems of current tests. This proposal entails the specification of measurement objectives, the development of test items for each objective, the development of test items to measure related objectives, and the provision of both a score and a score interpretation for each objective. Responses to potential criticisms of this approach, including its reliability and its usefulness to teachers, are made. It is concluded that tests should be evaluated in terms of the quantity, quality, and the cost of the information they provide. (PR)