Instructional improvement within the context of criterion-referenced and norm-referenced tests is described. Such categories overemphasize test interpretation rather than design characteristics of achievement tests. Data from most measurement situations may be reported or interpreted either according to criterion- or norm-referenced standards. How the test is developed and what it represents is of critical importance. The paper proposes alternative conceptualizations of test design: construct-referenced, objectives-referenced and domain-referenced. Using student data, the teacher needs to identify deficiencies in achievement, possible explanations, and remedies and to put the remedies into operation. An analysis of the utility of each test type results in the appraisal that domain referenced tests provide the most information for teachers and therefore are the most desirable as data sources for instructional improvement. However, because of lack of knowledge about instruction, poor training in available instructional principles, and lack of resources to encourage changes in instructional habits, it is concluded that instructional improvement, even if measurement considerations were satisfied, is not imminent. (Author/DJ)