The author contends that the crucial variable distinguishing two types of tests is the way in which the content domain is specified, rather than the use of norm-vs. criterion-referenced interpretations. Because some modes of content specification lead to specified domain and mastery interpretations, and some lead to open domains, the author recommends a classification system for distinguishing between different types of tests, especially with respect to the kinds of interpretations that can be derived from their scores. Tests are discussed in terms of summative and formative functions in the evaluation of learning. Five modes for specifying test content are outlined. These include the content/process matrix, theoretical construct, criterion sampling, objectives-based, and formal item generation rules. In addition, domain-referenced, criterion-referenced, and norm-referenced methods of test score interpretation are outlined. In a discussion of content specification mode/function combinations, content/process is seen as the most flexible specification modes. Theoretical constructs used as specification modes are primarily useful for the formative function of diagnosis and the summative function of selection/prediction. Overall, modes for generating specified content domains are described as particularly appropriate for the two formative functions of placement and diagnostic/progress, and for the summative function of certification/crediting. (JAC)