99 Ways to Obtain a General Factor of Intelligence—And They Are Not All Created Equal

Abstract

This paper examines whether general intelligence (g) factors derived from different test batteries are equivalent. There are three views regarding the equivalency of g-factors: (1) “indicator indifference” claims that test content is irrelevant as long as g loadings are identical and that single tests can be adequate indicators of g; (2) “complete dependence on test composition” claims that general factors are completely dependent on the tests from which they are extracted; and (3) an intermediate stance that emphasizes the importance of the diversity and comprehensiveness of cognitive test batteries from which g is obtained. The present study evaluates these competing views by analyzing g-factor correlations across all combinatorically possible combinations of subtests from the Woodcock–Johnson Tests of Cognitive Abilities V. Results showed strong correlations among g-factors across both one-factor and hierarchical models, increasing with the number of subtests or broad abilities included. Most g-factors closely matched the g-estimate obtained from all available subtests (r > .9). Low correlations were mainly tied to the overrepresentation of processing speed (Gs) in small test sets, highlighting the impact of content coverage. Overall, results support the intermediate view: reliable g-estimates require broad, balanced batteries covering at least three broad abilities.