These four characteristics inherent in criterion-referenced tests form the central theme of this paper: (1) The classes of behaviors that define different achievement levels are specified as clearly as possible before the test is constructed; (2) Each behavior class is defined by a set of test situations (that is, test tasks) in which the behaviors can be displayed in terms of all their important nuances; (3) Given that the classes of behavior have been specified and that the test situations have been defined a representative sampling plan is designed and used to select the test tasks that will appear on any form of the test; and (4) The obtained score must be capable of expressing objectively and meaningfully the individual's performance characteristics in these classes of behavior. The focus of this paper is on the development of criterion-referenced tests having these four properties and some associated technical problems that are encountered. Solutions for these technical problems are not readily available nor immediately generalizable to all curricular areas for which criterion-referenced tests might be desired. Attempts are made, therefore, to specify procedures that will be useful to the practical developer until the technical problems are solved. (RC)