A major aspect of the problem of evaluation concerns the measurement of educational outcomes. Such measurement involves, first of all, an inference regarding what is being learned, retained, or transferred. A description is given of the operations which underlie such inferences. It is contended that educational measurement needs to concern itself with the distinction among a number of classes of learning outcomes, particularly verbal sequences, concepts and principles. Primary criteria in such measurement are considered to be distinctiveness and freedom from distortion. Procedures of laboratory psychologists are described in achieving these criteria in the measurement of such outcomes as conditioned responses, chains and multiple discriminations. Analogous procedures need to be employed in designing items to measure concepts and principles with distinctiveness, and free from distortion. One technique described is two-stage measurement, in which the simpler outcome is measured first, then the more complex. (Author)