Slope-Controlled Performance Testing

Abstract

Cognitive-ability tests, though promising in other respects, generally show pronounced practice effects and have weak test-retest reliabilities. One reason for the low reliabilities appears to be that practice effects themselves vary from individual to individual, so that subjects differ not only in the levels at which they are performing when testing ends but also in the slopes leading up to those levels. Since slope of the performance curve late in practice has been shown to affect performance at reacquisition (retest), uncontrolled variation in slope may lower test-retest reliability. A possible approach to this problem is experimentally to control slope during testing so that all subjects are improving at roughly the same rates when testing ends. The expected effect is that, with inter-subject differences in slope controlled, the temporal stability of cognitive-ability tests will improve. If temporal stability improves, however, predictive validities ought also to improve.