Limited Internal Comparability of General Intelligence Composites: Impact on External Validity, Possible Predictors, and Practical Remedies

Abstract

Research on comparability of general intelligence composites (GICs) is scarce and has focused exclusively on comparing GICs from different test batteries, revealing limited individual-level comparability. We add to these findings, investigating the group- and individual-level comparability of different GICs within test batteries (i.e., internal score comparability), thereby minimizing transient error and ruling out between-battery variance completely. We (a) determined the magnitude of intraindividual IQ differences, (b) investigated their impact on external validity, (c) explored possible predictors for these differences, and (d) examined ways to deal with incomparability. Results are based on the standardization samples of three intelligence test batteries, spanning from early childhood to late adulthood. Despite high group-level comparability, individual-level comparability was often unsatisfactory, especially toward the tails of the IQ distribution. This limited comparability has consequences for external validity, as GICs were differentially related to and often less predictive for school grades for individuals with high IQ differences. Of several predictors, only IQ level and age were systematically related to comparability. Consequently, findings challenge the use of overall internal consistencies for confidence intervals and suggest using confidence intervals based on test–retest reliabilities or age- and IQ-specific internal consistencies for clinical interpretation. Implications for test construction and application are discussed.