Title

Investigating the determinants of cognitive ability test performance: Is "ei" more than meets the eye?

Date of Completion

January 1999

Keywords

Psychology, Industrial|Psychology, Psychometrics|Psychology, Cognitive

Degree

Ph.D.

Abstract

Classical True Score Theory (Nunnally, 1978) would suggest that Cognitive Ability Test (CAT) Performance is a function of the sum of true cognitive ability and measurement error. In this paper, it is proposed that there are determinants in addition to ability and error that systematically affect test performance. The proposed determinants of CAT performance are evaluated in a causal framework using structural equation modeling techniques. The determinants include test-taker motivation, dispositions, reactions to the test, prior testing experience, and cognitive ability. In addition, the hypothesis that racial subgroup differences on CATs may be partially explained by the role of the proposed determinants on CAT performance was tested. ^ Multiple-groups structural equation modeling was used to analyze test and survey data from 241 Black and 607 White employees from a hi-tech manufacturing organization who took a computerized test battery as part of a concurrent test validation study. The results show that the latent distribution of cognitive ability does not differ by race and that Blacks and Whites do not differ significantly on the other variables tested in this study that have an effect on CAT performance. Therefore, it is not clear from this research that Blacks and Whites differ on key latent variables that cause differences in mean group cognitive test scores as was proposed. ^ The results of this study do show that factors other than true cognitive ability significantly affect the measurement of CAT performance for both groups either directly (e.g., test-taking anxiety) or indirectly (e.g., belief in testing and prior computer experience) through an intervening variable. ^