Self-adapted testing: Improving performance by modifying tests instead of examinees

Abstract

This paper describes self-adapted testing and some of the evidence concerning its effects, presents possible theoretical explanations for those effects, and discusses some of the practical concerns regarding self-adapted testing. Self-adapted testing is a variant of computerized adapted testing in which the examinee makes dynamic choices about the difficulty of the items he or she attempts. Self-adapted testing generates scores that are, in contrast to computerized adapted test and fixed-item tests, uncorrelated with a measure of trait test anxiety. This lack of correlation with an irrelevant attribute of the examinee is evidence of an improvement in the construct validity of the scores. This improvement comes at the cost of a decrease in testing efficiency. The interaction between test anxiety and test administration mode is more consistent with ainterference theory of test anxiety than a deficit theory. Some of the practical concerns regarding self-adapted testing can be ruled out logically, but others await empirical investigation.