Noting that in recent years testing has fallen into disrepute, this paper demonstrates how the construction, use, and validation of tests, if adequately conceived, can increase the level of knowledge about the intricate nature of the reading processes. The first section of the paper discusses the measurement process and the process of test validation, and examines reasons why tests have often failed as instruments of scientific inquiry in reading. In addition, the section describes the Assessment of Reading Processes (ARP), a construct-referenced measure that views reading as a constant exchange between reader and text. The second section of the paper presents a theory of assessment that defines (1) the limits of the reading processes, (2) how such covert mental operations may be inferred from the reading product, and (3) rules for the generation of the structure and content of assessment items. Other topics discussed in this section include the assessment situation, the reader-text interaction, and the cognitive processes involved in reading. The third section offers a model of construct-referenced test construction and discusses each of the components of such a model, while the last section discusses the construction of the ARP. (FL)