Correlations among cognitive abilities are lower for higher ability groups

Abstract

Previous research demonstrates that correlations among IQ tests are lower when estimated using higher ability groups ( Detterman & Daniel, 1989; Lynn, 1990 ). However, this phenomenon has only been documented using individually administered measures of intelligence and attempts to extend the demonstration to other specialized abilities have failed ( Detterman, 1993 ). This study divides the 1980 Armed Services Vocational Aptitude Battery (ASVAB) weighted norming sample into five ability levels with similar levels of variance. Analyses show that ASVAB tests are less correlated within higher ability groups provided that the scales used to define the groups are psychometrically sound: For three highly skewed ASVAB tests, a ceiling effect prevents this phenomenon; for the remaining seven tests, the phenomenon replicates; and the magnitude of the effect is proportional to the skewness of the scale, r = −.85. These findings support the assertion that cognitive abilities are less correlated in higher ability groups, imply that greater classification effects can be associated with more intelligent groups, and qualify the use of the multivariate correction for restriction of range.