It is usually assumed that the procedures followed in developing criterion-referenced (CR) educational achievement tests ensure their content validity. This assumption is reasonable for the class of "domain-specified" CR tests, in which the item domain is completely specified by item generation rules. However, most CR tests are of the "content-specified" type, which are developed from more general content specifications which describe but do not determine the item domain. Similar considerations of validity apply for content-specified CR tests as for norm-referenced (NR) tests which are developed in a similar manner. Content validity is even more important for CR tests than for NR tests, because there is no normative group and seldom is there other evidence of the validity of the test. The CR test developer attempts to develop a test which represents the set of criterion performances which the test purports to measure, but it is actually very difficult to determine how well this goal has been achieved. The best assurance of CR test validity is careful adherence to high standards for test development. (Author/MV)