Reconsidering the search for alternatives to general mental ability tests

Abstract

Cognitive ability tests that measure general mental ability (g-tests) are among the best predictors of academic, training, and job performance. One disadvantage of g-tests is the potential for adverse impact due to subgroup differences on general mental ability (g). For many years, psychologists have searched for high-validity low-adverse impact alternatives to traditional g-loaded cognitive ability tests (g-tests). This paper explores the mathematical possibility of developing such a test based on the known characteristics of g-tests. It was discovered that superior replacements to g-tests cannot mathematically exist. This is due to the fact that adverse impact and subgroup differences occur primarily on g rather than the specific factors and unique variance that cognitive ability tests measure. The reliable non-g variance in most g-tests is too small to offset the subgroup differences in g-test scores that is attributable to g.