Validity, Types of

The meaning of validity has changed often over the past century (see Kane, 2001). Originally, it was related to whether a test measured what it purported to measure, but in the mid-1950s, the usage of the concept became more complex as different types of validity were identified. Thus, it came to refer to face, content, criterion, and concurrent validity. (Face validity refers to how well the test appears, “on its face,” to measure what it is claimed to measure. Content validity refers to how well the test instrument covers or samples from all aspects or content of the entity or domain it is attempting to measure—for example, to what degree does a test of mastery of a science curriculum adequately sample from all the content covered in that course. Criterion and concurrent validity are closely related and can be thought of as referring to how well the test correlates with a measure of the criterion taken at the same time and which has already been validated—for example, how well does a new way of measuring IQ [intelligence quotient] produce results that highly correlate with the standard measure of IQ.) Validity was also distinguished from reliability, which is the degree to which the test produces the same result on subsequent administrations (therefore, a test that is not valid can nevertheless be reliable). This entry focuses on the developments that subsequently have taken place with regard to the conception of validity and on the implications for users of tests.

These conceptions of validity relied heavily on providing evidence for the quality of the items used in the test, the scores, the scoring, and the test structure. Over the past half-century, however, an important change in emphasis has gradually occurred—discussions of validity became more concerned with the quality of a test score as a basis for making defensible interpretations. That is, the great shift has been from providing evidence of validity about the test to providing validity evidence about the way the results from the test are used—in short, the focus now is on the validity of the interpretations or inferences from the test. As pioneers in this area, Cronbach and Meehl put it, as long ago as 1955, “One does not validate a test, but only a principle for making inferences” (p. 297). Messick (1989) has been most forceful in providing a unified approach to the notion of validity. He claimed,

An important implication of the view of validity expounded by Messick is that it is vital that [Page 830]evidence is provided about how users make inferences and take actions on the basis of test scores, and it also highlights the importance of determining whether the test and report developer can provide evidence for the adequacy and appropriateness of these interpretations. It should be clear that validity is not a “yes” or “no” matter; rather, it can be assessed via the degree to which the accumulated evidence supports a particular test use or interpretation. No one form of evidence can suffice; instead, the evidence needs to be multiple and aimed at defending the quality of the interpretations made on the basis of the test score (see Cronbach, 1988). The key, however, is that professional judgment is required to determine the forms of evidence that are most appropriate in a given situation and to judge the adequacy of the support for the intended purpose. In some cases, more rigor may be needed, such as when high-stakes decisions are being made.

...