Gene x SES Interaction and Age-related Increases in the Heritability of Cognitive Ability

Background: The proportion of variance in cognitive ability due to genetic factors fluctuates across both age and environmental context. Numerous behavior genetics studies have demonstrated that the heritability of cognitive... [ view full abstract ]

Background: The proportion of variance in cognitive ability due to genetic factors fluctuates across both age and environmental context. Numerous behavior genetics studies have demonstrated that the heritability of cognitive ability increases substantially as age rises. Gene x socioeconomic status (SES) interaction has also been observed in most U.S. samples, where individuals reared in higher SES environments show greater heritability of cognitive ability than lower SES peers. To date, these phenomena have been studied largely in isolation from each other, and the specific mechanisms that underlie them remain poorly understood. In this study, we propose that both G x SES interaction and age-related increases in the heritability of cognitive performance are driven by the same process: divergence in dizygotic (DZ) twins. Using longitudinal cognitive data from middle childhood and early adolescence, we tested whether 1) DZ twins become more phenotypically distinct more quickly than monozygotic twins with rising age, resulting in increased heritability; and 2) the rate of this divergence is modified by SES.

Methods: Hollingshead SES and longitudinal WISC data were drawn from the Louisville Twin Study (LTS). Data collection for the LTS ran from 1957 until the late 1990s, generating arguably the most comprehensive twin data on childhood cognitive development ever assembled. Twin pairs were all from the Louisville, Kentucky area, and were included in the current analyses if cognitive and SES data at ages 7, 8, 9, 12, and/or 15 years were available for both twins. We analyzed data from 566 twin pairs in total (282 monozygotic (MZ), 284 DZ; 236 same sex female, 210 same sex male, 120 opposite sex). 80.04% of the sample participated in data collection at three or more ages. We modeled divergence with a latent growth curve (LGC) model. Latent intercept and slope factors were created for the between-pair and within-pair variances in cognitive ability. Divergence was quantified by the within-pair slope and the within-pair covariance between intercept and slope. We tested whether these parameters differed between MZ and DZ twins. To test whether the magnitude of within-family divergence in cognitive ability varied as a function of SES, we compared the -2 log likelihoods of four models. The first included the base LGC model plus linear main effects of SES on between- and within-pair intercept and slope factors, moderation of the within-pair slope variance, and moderation of the covariance between the within-pair intercept and slope factors. In the second model, the moderation of the within-pair slope variance was constrained to 0. In the third model, moderation of the within-pair intercept and slope correlation was constrained to 0. Finally, in the fourth model, the moderation of both the within-pair slope variance and the covariance of the within-pair intercept and slope were constrained to 0. All twin analyses were performed using OpenMx with full information maximum likelihood estimation and NPSOL optimization.