Individual differences in intelligence significantly predict various life outcomes, including life satisfaction, mortality, and educational achievement. The sources of these differences have been a subject of debate, with research highlighting the roles of genetics and environmental factors. Schooling is a crucial environmental factor substantially impacting children's intelligence. Longitudinal studies, controlling for prior intelligence, and studies evaluating the effects of compulsory schooling policy changes, demonstrate schooling's positive impact on cognitive abilities. Regression discontinuity designs, exploiting age-based grade allocation, further support this finding. While the effect of schooling on intelligence is well-established, its interaction with children's environments and genetic predispositions remains unclear. Socioeconomic status (SES), encompassing household income, parental education, and neighborhood quality, is a significant environmental factor influencing intelligence. Research suggests that SES can exacerbate existing intelligence differences. However, this research often overlooks the role of genetics, as parents transmit both genes and environments to their children. Studies using polygenic scores (PGS), which aggregate the effects of numerous genetic variants, offer a way to incorporate genetic information into this research. A multi-trait cognitive polygenic score (cogPGS) has been shown to predict a significant portion of variance in cognitive performance and correlates with SES. Gene-by-environment (GE) interplay is often suggested to explain the coexistence of high heritability and malleability in intelligence. However, studies investigating GE interplay using PGS have yielded inconsistent results, particularly regarding schooling's role. This study aims to examine the unique contributions of schooling, SES, and cogPGS to different domains of intelligence (crystallized intelligence, fluid intelligence, and working memory) and to investigate whether schooling interacts with pre-existing genetic and environmental inequalities.

Decades of research have established a strong association between intelligence and genetics (heritability estimates around 0.6), but environmental factors also play a significant role. Schooling is a crucial environmental factor with a large impact on intelligence in children, demonstrated by longitudinal studies controlling for prior intelligence, and studies examining the cognitive effects of policy changes in compulsory schooling. Regression discontinuity designs, leveraging age-based grade cutoffs, provide further evidence that schooling impacts intelligence. Existing literature also highlights the role of socioeconomic status (SES) in shaping intelligence differences, with some research suggesting SES widens existing disparities. However, a critical gap in this research is the neglect of genetic influences, since parents transmit both genes and environments. Recent advancements in genome-wide association studies (GWAS) and the development of polygenic scores (PGS) have made it possible to incorporate genetic information into analyses of intelligence and its determinants. A multi-trait cognitive polygenic score (cogPGS) predicts cognitive performance and correlates with SES, but its unique contributions to different intelligence domains and interactions with schooling are largely unknown. Prior studies exploring gene-environment interplay using polygenic scores primarily focused on SES as the environmental factor, with inconsistent findings. Educational attainment research also shows inconsistencies in interactions between genetics and SES. Because educational achievement is a broader concept than intelligence, and these findings are not directly applicable to intelligence and its domains, this study employs a regression discontinuity approach to investigate the interaction of schooling with cogPGS and SES on intelligence.

This study utilized data from the Adolescent Brain Cognitive Development (ABCD) study, encompassing 6567 children (mean age 9.88, range 8.92–11.00) in grades 3-5. Socioeconomic status (SES) was defined as the first component of a probabilistic PCA, combining household income, parental education, and neighborhood quality. A multi-trait cognitive polygenic score (cogPGS) was calculated using PRSice-2, incorporating data from a large genome-wide association study on educational attainment, mathematical ability, and general cognitive ability. Cognitive measures included working memory (WM), fluid intelligence (fIQ), and crystallized intelligence (cIQ) from the NIH toolbox cognition battery. A regression discontinuity design was employed to isolate the effect of schooling, controlling for age, sex, cogPGS, SES, and ancestry-based principal components. Regression models were used to examine the unique effects of schooling, SES, and cogPGS on each cognitive domain and to assess interaction effects. A hierarchical Bayesian mixed-effects model was used for post-hoc Bayesian null hypothesis testing, employing region of practical equivalence (ROPE) boundaries. A post-hoc sibling analysis (392 families, n = 792) was conducted to disentangle within-family and between-family effects of cogPGS. A post-hoc confirmatory factor analysis was performed to extract a general intelligence (g) factor score, and analyses were repeated using this g factor. Finally, a post-hoc analysis was conducted on a subset of individuals of European ancestry to address potential limitations of cogPGS accuracy in non-European populations. The data was processed using R, with packages including lme4, lmerTest, brms, and Lavaan.

The study found significant, independent effects of schooling, cogPGS, and SES on cIQ, fIQ, and WM. Schooling had a substantial effect on all three cognitive domains (cIQ: β = 0.13; fIQ: β = 0.10; WM: β = 0.09). One year of schooling contributed 0.22 SD to cIQ, 0.14 SD to fIQ, and 0.14 SD to WM. The effect of one year of schooling was larger than that of one year of chronological age, especially for WM (ratio = 2.2). Both cogPGS and SES had significant, independent effects on all three cognitive domains, with larger effects for cIQ than for fIQ or WM. The effects of parental education and income were similar in size and larger than the effects of neighborhood quality on cognitive ability. A sibling analysis revealed that the within-family effect of cogPGS on cIQ was roughly half the between-family effect, suggesting passive genotype-environment correlations. There were no significant interactions between schooling and cogPGS or SES for any of the cognitive domains. This null result remained consistent even after adding principal component interactions. Bayesian analysis indicated that effects smaller than 0.02 SD could not be ruled out, but most interaction effects were less than 0.05 SD. Post-hoc analysis using a general intelligence (g) factor yielded similar results. A post-hoc analysis restricting the sample to individuals of European ancestry showed consistent results, albeit with slight changes in effect sizes for cogPGS and SES. The two years of schooling studied had a larger effect than the lifetime effects of SES or cogPGS.

This study demonstrates the substantial and independent effects of schooling on intelligence, with two years of schooling having a greater impact than lifetime effects of SES or cogPGS differences. The lack of significant interactions between schooling and either SES or cogPGS suggests that schooling does not exacerbate or mitigate pre-existing socioeconomic or genetic inequalities in cognitive development. While the null results for interactions warrant cautious interpretation due to potential limitations in detecting small effects, the findings challenge the notion that schooling either amplifies or diminishes existing inequalities. The substantial independent effects of schooling highlight its importance in cognitive development. These results contribute to ongoing debates about the relative contributions of nature and nurture to intelligence and the potential for educational interventions to reduce inequality. Future research should address the limitations of this study, focusing on larger samples, longer schooling periods, and exploring alternative methods for assessing gene-environment interactions.

This study provides strong evidence for the significant positive effects of schooling on intelligence. Importantly, the findings indicate that schooling does not appear to alter the existing rank order of children's intelligence based on socioeconomic status or genetic predispositions. While limitations remain concerning the detection of small interaction effects, the results suggest the need for continued efforts to ensure equitable access to quality education for all children. Future research could explore potential cumulative effects of schooling across longer timeframes and investigate more nuanced gene-environment interactions using different methods.

The study's limited range of schooling (grades 3-5) may constrain the detection of small but potentially meaningful cumulative interaction effects. The cogPGS, derived from a predominantly European-ancestry sample, may limit generalizability to other populations. The exclusion of children with missing DNA data resulted in a sample with a higher-than-average SES, potentially limiting the detection of interaction effects at the lower tail of the SES distribution. The study's power to detect small interaction effects, especially for schooling, might be limited. Lastly, the interpretation of findings from multi-trait GWAS should consider the potential influence of one supplementary phenotype driving the results.