Validity

Test Validity

Test validity, or the validity of an assessment, measure, or inventory, exists as a matter a degree. Test validity refers to the extent to which evidence and theory support the accuracy of the interpretation of the test score. Historically, test validity has been viewed as consisting of three different types of validity: construct validity, content validity, and criterion-related validity. This tripartite view of validity has been challenged, however, by other researchers, who suggest that there is one validity and several other sources of evidence of that validity. In this view, construct validity is the only true test validity, and content and criterion-related validity (as well as the other types of validity discussed below) are sources of evidence of construct validity. Although this view is considered to be a more accurate understanding of validity, the tripartite view is still commonly taught.

In both views, construct validity is central. Construct validity can be defined by the extent to which the test measures the construct that it claims to measure. For instance, those who use the body mass index claim it to be a measure of health, specifically healthy weight. The extent to which the body mass index accurately describes whether or not someone is at a healthy weight reflects the measure’s construct validity. Using a different example, if an individual is diagnosed with attention-deficit/hyperactivity disorder, but the test actually assessed the personality trait extraversion, then the inference based on the test results, that the individual has attention-deficit/hyperactivity disorder, is not valid.

Determining the extent of construct validity starts with building a nomological network, which describes how the construct fits into a network with other constructs. For instance, researchers must decide whether attention-deficit/hyperactivity disorder is expected to be positively or negatively related, or unrelated, to extraversion. Once this nomological network is developed, the extent to which the measure shows convergent and discriminant validity can be assessed. Convergent validity exists when a measure of one construct is positively correlated with a different measure of the same, or a similar, construct, identified by the proposed nomological network, whereas discriminant validity exists when a measure of the construct is negatively related or unrelated to constructs which are not in the proposed nomological network.

Content validity is the extent to which the content, or specific items, on a measure or assessment reflect the content that theoretically should be on [Page 2290]the assessment. In other words, content validity is primarily concerned with relevance and representativeness of the test content. For instance, in determining whether children are developing their language skills at an appropriate rate, it is useful to assess several aspects of language, not just vocabulary. Only assessing vocabulary, when the goal is to assess language acquisition in general, would lead to construct deficiency or underrepresentation. Similarly, although motor skills are also an essential aspect of development, they are not related to language development specifically. Therefore, assessing motor skills would be considered construct irrelevant. To ensure that a measure contains items which are relevant and representative of the construct being measured, the test should typically be created by, or at least reviewed by, several experts who are knowledgeable about the construct that is being assessed.

...