Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
1
From genetic disposition to academic achievement: The mediating role of
non-cognitive skills across development.
Quan Zhou1, Wangjingyi Liao1, Andrea G. Allegrini2,3, Kaili Rimfeld3,4, Jasmin Wertz5, Tim Morris6,
Laurel Raffington7, Robert Plomin3, Margherita Malanchini1,3
1 School of Biological and Behavioural Sciences, Queen Mary University of London, London, UK
2 Division of Psychology and Language Sciences, University College London, London, UK
3 Social, Genetic & Developmental Psychiatry Centre, Institute of Psychiatry, Psychology &
Neuroscience, King's College London, London, UK
4 Department of Psychology, Royal Holloway University of London, London, UK
5 School of Philosophy, Psychology and Language Sciences, The University of Edinburgh, Edinburgh,
UK
6 Centre for Longitudinal Studies, Social Research Institute, University College London, London, UK
7 Max Planck Research Group Biosocial – Biology, Social Disparities, and Development; Max Planck
Center for Human Development, Berlin, Germany
Abstract
Genetic effects on academic achievement are likely to capture environmental, developmental, and
psychological processes. How these processes contribute to translating genetic dispositions into
observed academic achievement remains critically under-investigated. Here, we examined the role of
non-cognitive skills—e.g., motivation, attitudes and self-regulation—in mediating education-
associated genetic effects on academic achievement across development. Data were collected from
5,016 children enrolled in the Twins Early Development Study at ages 7, 9, 12, and 16, as well as
their parents and teachers. We found that non-cognitive skills mediated polygenic score effects on
academic achievement across development, and longitudinally, accounting for up to 64% of the total
effects. Within-family analyses highlighted the contribution of non-cognitive skills beyond genetic,
environmental and demographic factors that are shared between siblings, accounting for up to 83% of
the total mediation effect, likely reflecting evocative/active gene-environment correlation. Our results
underscore the role of non-cognitive skills in academic development in how children evoke and select
experiences that align with their genetic propensity.
Introduction
Academic achievement, commonly measured as school grades during childhood and adolescence, is
associated with a range of positive life outcomes, including better physical and mental health1,2 and
higher income3–5. Research indicates that genetic differences between people partly explain variations
in academic achievement, a concept known as heritability6. Twin studies, which estimate the relative
contribution of genetic and environmental factors by comparing identical and fraternal twins, have
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
2
shown that 40 to 60% of the differences in educational attainment and academic achievement can be
attributed to genetic factors7,8.
DNA-based methods have also quantified academic achievement, estimating heritability at 20 to 30%
during childhood and adolescence7,9,10. Another way of capturing education-associated genetic effects
is to construct a polygenic score (PGS)11. Polygenic scores aggregate the effects of numerous genetic
variants identified in large-scale genome-wide association studies, creating a composite index that
captures known genetic effects on a trait12. Education-associated polygenic scores have been found to
account for 12 to 16% of the variance in adult educational attainment (years of schooling)13. In
addition, a polygenic score combining education14 and cognition-associated genetic variants was
found to account for 15% of the variance in academic achievement at the end of compulsory
education in a UK-based sample15, a figure comparable to other well-established predictors of
academic achievement, such as family socioeconomic status16.
Genetic effects on academic achievement likely encompass a combination of environmental,
developmental, and psychological processes. To explore this further, two recent studies investigated
how family environments mediate these genetic influences. The first study showed that, beyond
socioeconomic status, multiple aspects of the family environment – such as supportive parenting,
household chaos, and TV consumption – mediated polygenic score associations with academic
achievement throughout compulsory education17. By employing a sibling-difference design to
distinguish genetic associations shared between siblings (between-family effects) and those unique to
each child (within-family effects), the study found that family environments were predominantly
correlated with between-family genetic effects17. These effects likely reflect passive gene-
environment correlation, suggesting that children, by growing up with their biological parents, are
exposed to environmental experiences that correlate with their genetics18. The second study found that
mothers’ educational attainment polygenic scores were associated with their children’s educational
outcomes, beyond the children's own genetics. This ‘genetic nurture’19 pathway from mother’s
genetics to her children’s attainment is thought to capture environmental processes that covary with
parental genetic propensities20 and was mediated by cognitively stimulating parenting21. However,
environmental, developmental and psychological experiences other than family environments that
may translate genetic propensity into academic achievement remain under-investigated.
The current study investigates the role of non-cognitive skills in mediating education-associated
genetic effects on academic achievement across development. The term ‘non-cognitive skills’
describes a set of attitudes and characteristics that impact life outcomes beyond what cognitive tests
can measure and predict22. These skills encompass motivation, perseverance, mindset, learning
strategies, and social skills, and self-regulatory strategies23. Non-cognitive skills independently
contribute to educational success beyond cognitive ability. A study of 16-year-olds found self-efficacy
and personality traits significantly influenced academic achievement beyond cognitive ability24. Other
studies also link personality, self-regulation, and motivation to academic success25–28. More recently,
our research highlighted that the association between non-cognitive skills and academic achievement
substantially increased from ages 7 to 1629. Furthermore, previous research showed that greater self-
control and, to a lesser extent, interpersonal skills, partly explain the genetic link to educational
attainment30.
We explored the role of non-cognitive skills in mediating education-genetic effects on academic
achievement through five key questions: First, do non-cognitive skills mediate polygenic score
associations with academic achievement across compulsory education? Second, can mediation effects
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
3
be captured even when separating education-associated genetics into a cognitive and a non-cognitive
component29,31 (i.e., are education-associated cognitive and non-cognitive genetic effects associated
with academic achievement via partly distinct non-cognitive mechanisms?)? Third, and importantly,
do non-cognitive skills mediate polygenic score-academic achievement associations longitudinally?
Fourth, do non-cognitive skills mediate polygenic score associations with academic achievement even
when limited to differences within pairs of siblings? And lastly, is the mediating role of non-cognitive
skills robust when controlling for other factors known to be related to both academic achievement and
non-cognitive skills, namely general cognitive ability and family socio-economic status?
To address these questions, we employed a design with unique features, including the measurement of
several non-cognitive skills across development, such as academic motivation, academic interest, and
self-regulation at age 7, 9 ,12 and 16 (see Methods). Data on non-cognitive skills were obtained from
multiple raters, including teacher, parents, and the children themselves. Furthermore, we used a
sibling difference design allowing to partly adjust for shared sibling processes, such as passive gene-
environment correlation17,32, demographic between-family factors, population stratification33, and
assortative mating34. These preregistered analyses (https://osf.io/vmpf7/) provide a detailed
exploration of how non-cognitive skills contribute to translating genetic disposition into observed
differences between children in academic achievement across compulsory education.
Results
Creating latent dimensions of non-cognitive skills
Several measures of non-cognitive skills were collected at different ages, with ratings provided by
parents, teachers, and the twins themselves. Following our previous work, two latent dimensions of
non-cognitive skills were created using confirmatory factor analyses29: education-specific non-
cognitive skills (including attitudes towards school, self-perceived academic ability, academic interest
and curiosity) and self-regulation (encompassing behavioural and emotional regulation). See Methods
for a detailed description.
Correlations between polygenic scores, non-cognitive skills and academic achievement
Descriptive statistics are presented in Supplementary Note 1 and Supplementary Table 1. All
variables were normally distributed; therefore, no transformations were applied prior to analyses.
We first analysed correlations of polygenic scores constructed from prior genome-wide association
studies of educational attainment (EA14), cognitive (Cog) and non-cognitive (NonCog) skills,
operationalised as genetic variants associated with education after accounting for genetic effects
associated with cognitive performance29, with academic achievement and the latent non-cognitive
dimensions rated by parents, teachers and self across compulsory education. These analyses were
performed on unrelated individuals by selecting one twin from each pair randomly to account for
relatedness. The correlations between polygenic scores and academic achievement at all ages were
statistically significant, moderately sized (Supplementary Figure 1) and increased across
development, particularly for the EA polygenic score (r ranging from 0.25 at age 7 to 0.36 at age 16)
and the NonCog polygenic score (r ranging from 0.10 at age 7 to 0.22 at age 16; Supplementary
Figure 1). The correlations between polygenic scores and latent dimensions of non-cognitive skills
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
4
were modest and statistically significant, except for self-rated non-cognitive skills at age 9 (see
Supplementary Figure 2; Supplementary Table 2). The correlations between latent dimensions of
non-cognitive skills and academic achievements were positive, moderately sized and statistically
significant across all ages (r ranging between 0.13 and 0.63) (Supplementary Figure 2;
Supplementary Table 2). Moderate and positive correlations were also observed when considering
individual measures of non-cognitive skills, including academic interest, self-perceived ability, grit,
and curiosity (see Supplementary Table 3).
Non-cognitive skills mediate polygenic score effects on academic achievement
We applied mediation models to examine the extent to which latent dimension of non-cognitive skills
(i.e., education-specific non-cognitive characteristics and self-regulation) mediated the association
between genetic disposition towards education, measured using the educational attainment polygenic
score, and academic achievement across compulsory education. Figure 1 (top panel) depicts the
results of the mediation models for the educational attainment polygenic score. Indirect (mediated)
effects for self-regulation were developmentally stable, similar across different raters, and small in
magnitude, ranging from ß = 0.01 to ß = 0.03. In contrast, indirect (mediated) effects for education-
specific non-cognitive skills increased across compulsory education, varied across raters, and were
larger in magnitude. When considering self-reported measures, indirect (mediated) effects increased
from zero at age 9, to ß = 0.02 [95% CI, 0.01, 0.04] at age 12 to ß = 0.08 [95% CI, 0.06, 0.11] at age
16. The indirect (mediated) effects of education-specific non-cognitive skills at age 9 were larger for
parent-rated (ß = 0.05 [95% CI, 0.02, 0.07]) and teacher-rated measures (ß = 0.13 [95% CI, 0.10,
0.15], (see Fig 1 top panel; Supplementary Table 4).
The two dimensions of education-specific non-cognitive skills and self-regulation also mediated the
cognitive and non-cognitive polygenic score-academic achievement association across compulsory
education. The strongest mediation effects were observed for teacher-rated education-specific non-
cognitive skills at age 9, which accounted for up to 48% of the total cognitive polygenic score
prediction, and up to 64% of the total non-cognitive polygenic score prediction (Fig 1 middle and
bottom panels, and Supplementary Tables 5 and 6 for results on cognitive and non-cognitive
polygenic scores obtained using a different model31).
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
5
Fig 1: Polygenic score associations with academic achievement mediated by latent dimensions of
education-specific non-cognitive skills and domain-general self-regulation. The length of each bar
represents the total effect (standardised beta coefficient) between the polygenic scores for educational
attainment14 (EA; top panel), cognitive29 (Cog, middle panel) and non-cognitive29 (NonCog; bottom panel) and
academic achievement at ages 7, 9, 12, and 16. The orange portion of each bar shows the indirect effect (i.e.,
mediated by latent factors of non-cognitive skills), while the remaining portion of each bar shows the direct
effect of polygenic scores on academic achievement (i.e., not mediated by each non-cognitive measure). The left
panels show mediation effects for education-specific non-cognitive skills (NCS), while the right panels show
mediation effects associated with self-regulation. * = p < 0.05, ** = p < 0.01 after applying FDR correction. See
model estimates in Supplementary Tables 4-6.
A detailed analysis of education-specific non-cognitive measures as mediators of polygenic
score effects on academic achievement
We conducted a more detailed set of analyses to investigate the role played by specific non-cognitive
measures in the association between polygenic scores and academic achievements (Fig 2 and
Supplementary Tables 7-9). Mediation effects were stronger for student-centred non-cognitive
variables such as academic interest, academic perceived ability, self-concept and curiosity, as
compared to measures capturing broader aspects of the learning environment, such as classroom
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
6
satisfaction and peer support (see Fig 2). In line with the results we obtained when considering latent
non-cognitive dimensions, the mediating role of parent-rated and teacher-rated non-cognitive
measures at age 9 was stronger than what we observed for self-rated non-cognitive measures.
Although for self-rated non-cognitive measures, mediation effects increased developmentally, for
example, for academic self-perceived ability and academic interest (Fig 2).
By the end of compulsory education, the mediating role of self-rated non-cognitive measures was
substantial. For example, the indirect (mediated) effects of academic self-concept and maths interests
measured at age 16 accounted for 45% and 32% of the association between the EA polygenic score
and academic achievement at age 16, respectively (left panel of Fig 2). Moreover, when we
partitioned polygenic score effects into cognitive and non-cognitive genetics, we observed that
academic self-concept and interest mediated both polygenic score associations with academic
achievement. For the Cog polygenic score-academic achievement association, mediation effects
accounted for 52% and 42% of the association, respectively (middle panel of Fig 2); while for the
NonCog polygenic score for 38% and 27%, respectively (right panel of Fig 2). Overall, mediation
effects observed for individual non-cognitive measures were highly consistent across the three
polygenic scores considered (Fig 2 and Supplementary Tables 7-9).
Fig 2: Direct and indirect effects of polygenic score associations with academic achievement as mediated
by individual education-specific non-cognitive skills (NCS). Variables labelled with (P), (T), and (S)
represent different informants: (P) = Parent-rated; (T) = Teacher-rated; (S) = Self-rated. Each dot represents the
effect size (standardized ß coefficient) for direct (red) and indirect (blue) polygenic score effects on academic
achievement at ages 9, 12 and 16. EA = educational attainment polygenic score14 (left panel), Cog = cognitive
polygenic score29 (middle panel), and NonCog = non-cognitive polygenic score29 (right panel). Error bars
around each dot indicate standard errors. * = p < 0.05, ** = p < 0.01 after applying FDR correction.
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
7
Non-cognitive skills mediate genetic disposition of academic achievement longitudinally
Given the consistent mediation effects of education-specific non-cognitive skills that emerged from
our analyses at the same developmental stage, we next examined time-lagged effects longitudinally,
such as whether non-cognitive skills measured at age 9 mediated polygenic score effects on academic
achievement at ages 12 and 16. We observed persistent mediation effects even in these time-lagged
analyses, although effects were slightly attenuated in some, but not all, cases (Fig 3). For example, the
role of self-rated non-cognitive skills at age 12 as mediator of the association between EA polygenic
score and academic achievement at age 16 remained substantial (right panel of Fig 3). Findings were
highly consistent when we examined Cog and NonCog polygenic score effects (see middle and
bottom panel of Fig 3 and Supplementary Table 10) and for individual non-cognitive measures
(see Supplementary Table 11).
Fig 3: Time-lagged mediating effects of Education-specific non-cognitive skills at age 9 (left panel) and
age 12 (right panel) on the polygenic score (PGS) predictions of academic achievement over development.
The colour represents the direct (red) and indirect (blue) effects (standardised beta coefficient) of the
educational attainment14 (EA; top panel), cognitive29 (Cog) and non-cognitive29 (NonCog; bottom panel)
polygenic scores on academic achievement at ages 9, 12, and 16. Error bars indicate the standard error around
the effect size. * = p < 0.05, ** = p < 0.01 after FDR correction.
Within-family mediation analyses
We applied multi-level mediation models (see Methods) to separate polygenic score associations into
within- and between-family effects and further disentangle the processes underlying the role of non-
cognitive skills in these associations. The number of sibling (DZ) pairs included in the analyses
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
8
ranged between 1,104 to 2,895, depending on the data collection wave and measures included.
Family-level intraclass correlations (ICCs) for latent non-cognitive factors ranged between 0.23 and
0.52 (see Supplementary Table 12), indicating moderate to substantial family-related effects on the
development of non-cognitive skills.
Consistent with previous research13,29,35, we found that, the associations between EA polygenic score
and academic achievement were attenuated when looking at differences within families. However,
young people’s non-cognitive skills remained significant mediators of the association between the
educational attainment polygenic score and academic achievement even when comparing siblings
(Fig 4 top panel and Supplementary Table 13). Siblings with a higher education-related polygenic
score showed on average greater academic achievement, and this association was partly accounted for
by higher levels of non-cognitive skills. Notably, our finding of a developmental increase in the
mediating role of self-reported non-cognitive skills replicated at the within-family level (see left
panel of Fig 5).
For parent-rated non-cognitive skills, mediation effects were mainly evident at the within-family level
rather than between-families (i.e., between-family mediation effect: ß = 0.02 [95% CI, -0.01, 0.05];
within-family mediation effect: ß = 0.10 [95% CI, 0.05, 0.14]), accounting for 83% of the total
mediation effects. Polygenic score associations with academic achievement that are unique to each
child, rather than shared by siblings, were more likely to be mediated parents’ perceptions of non-
cognitive profiles unique to each child. While this was observed for education-specific non-cognitive
skills, associations were different for self-regulation. When considering children’s unique self-
regulation profiles, most mediation effects were attenuated and did not reach significance in within-
family analyses. We observed a similar pattern of results when repeating mediation analyses
separating education-related genetics into cognitive and non-cognitive polygenic scores (middle and
bottom panel of Fig 4; Supplementary Tables 14 and 15). Results were also consistent when we
examined the mediating role of each education-specific non-cognitive skills individually
(Supplementary Table 16).
We further investigated whether within-family mediation effects could be observed longitudinally.
We found that time-lagged mediation effects for education-specific non-cognitive measures persisted
even when looking within families. For example, parent-rated education-specific non-cognitive skills
measured at age 9 mediated EA polygenic score associations with academic achievement at both age
12 (ß = 0.08 [95% CI, 0.03, 0.12]) and age 16 (ß = 0.08 [95% CI, 0.05, 0.11]). The effect size was
comparable to the effect size observed cross-sectionally in the association between the EA polygenic
score and academic achievement measured at age 9 (that is, ß = 0.10 [95% CI, 0.05, 0.14]). The full
results are presented in Supplementary Tables 17-19 and provide evidence for the long-lasting
mediating role of education-specific non-cognitive skills in the association between polygenic score
and achievement across compulsory education. When examining the effects of each education-related
non-cognitive skill individually, we found that parent-rated academic perceived ability and academic
interest evinced the strongest mediation effects (ß = 0.09 [95% CI, 0.04, 0.15] and ß = 0.08 [95% CI,
0.05, 0.11], see Supplementary Table 20).
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
9
Fig.4: Mediation effects of the non-cognitive latent factors in the polygenic score (PGS) predictions of
academic achievement separated into between- and within-family effects. The dots represent the effects of
the PGS predictions (standardised beta coefficient) of academic achievement at ages 7, 9, 12 and 16 partitioned
into between- and within-family effects. Education-specific non-cognitive skills: left panel. Self-regulation:
right panel. Educational attainment polygenic score (EA PGS)14: top panel. Cognitive polygenic score (Cog
PGS)29: middle panel. Non-cognitive polygenic score (NonCog PGS)29: bottom panel. * = p < 0.05, ** = p <
0.01 after applying FDR correction.
Adjusting for general cognitive ability (g) and family socio-economic status (SES)
Since previous research has shown that students’ non-cognitive skills correlate with students’
cognitive ability26,29, we tested the possibility that cognitive ability could account for the mediation
effects we observe. Specifically, we repeated our mediation analyses, including general cognitive
ability as a second mediator (see Methods). Results showed that non-cognitive skills remained
significant, although slightly attenuated, mediators of the polygenic scores-achievement associations
(Supplementary Figure 3 and Supplementary Table 21). Results were similar across compulsory
education and across the different polygenic scores (Supplementary Figures 4 and 5;
Supplementary Table 22). Non-cognitive skills also remained significant mediators after accounting
for family socio-economic status (Supplementary Figure 6 and Supplementary Table 23), with
similar results observed across polygenic scores and compulsory education (Supplementary Figures
7 and 8; Supplementary Table 22).
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
10
Discussion
While it is well-established that genetic differences between students partly contribute to their
differences in academic achievement6,7, how genetic differences come to manifest as differences in
academic achievement remains under-investigated. To address this major gap in our knowledge, the
present study provides an in-depth examination of the role that non-cognitive skills (e.g., academic
motivation, attitudes and goals) play in the association between genetic disposition and observed
academic achievement across compulsory education. Our analyses revealed three key findings. First,
non-cognitive skills, particularly those skills closely linked to learning (e.g., academic interest,
curiosity, and academic self-perceived ability), play a critical role in translating education-associated
genetic dispositions into observed academic achievement. Second, our time-lagged analyses showed
that the mediating role of non-cognitive skills persisted across development. Third, these mediation
effects could also be observed when examining differences between siblings. Mediation effects were
small yet consistent and robust to adjusting for family socio-economic status and general cognitive
ability, highlighting the independent contribution of non-cognitive skills to academic success.
Our cross-sectional analyses showed that the latent dimensions of education-specific non-cognitive
skills and self-regulation both mediated the associations between education-related genetic disposition
and academic performance. However, while the role of self-regulation was stable across compulsory
education, the mediating role of education-specific non-cognitive skills increased developmentally,
pointing to the growing importance of students’ perceived non-cognitive profiles and experiences in
their academic journeys. This developmental increase is consistent with the possibility that, as they
grow up, children become more aware of their aptitudes and appetites towards learning. As they gain
greater self-awareness and autonomy, students might become increasingly more able to shape their
environmental contexts in ways that allow them to cultivate these non-cognitive skills and, in turn,
foster their academic performance36.
The magnitude of mediation effects also differed depending on raters. Teacher reports of education-
specific non-cognitive skills evinced the strongest effects, followed by parent reports, and self-reports.
These stronger effects may reflect rater bias, as teachers evaluated both students’ performance and
their learning-related attitudes37. However, they may also reflect the greater expertise and objectivity
that characterise teachers’ assessments38. This is supported by our finding that teacher-rated non-
cognitive skills had long-lasting mediation effects in our time-lagged analyses, since different teachers
assessed achievement at different ages. Teachers have first-hand experience of how children express
their attitudes and appetites within academic settings and might, therefore, be able to capture students’
non-cognitive competencies related to learning more accurately39.
With a set of fine-grained analyses, we investigated which aspects of non-cognitive skills played a
greater role in mediating the link between genetic disposition and academic achievement across
development. First, we observed that student-centred academic attitudes and perceptions, such as
academic interest, curiosity, and perceived ability, evinced stronger mediation effects as compared to
perceptions of the learning environment that were external to each individual, such as perception of
the classroom or the learning environments. These findings align with previous work, which
emphasised the critical role of intrinsic, student-cantered non-cognitive skills (e.g., intellectual
curiosity and academic self-concept) in gene-environment transactions36,40, driving students to
actively engage in academic settings beyond what environmental factors alone can achieve41. This
might be particularly useful knowledge for interventions aimed at fostering learning. Interventions
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
11
aimed at enhancing students' academic self-concept, interests, and motivations, such as inquiry-based
learning42, positive feedback43, and goal-setting44, might be particularly effective.
We examined whether the mediating role of non-cognitive skills was more strongly linked to
cognitive or non-cognitive genetics. We found that mediation effects were comparable when
separating education-associated genetic effects into cognitive and non-cognitive genetics.
Interestingly, for some student-centered non-cognitive skills such as academic interest and academic
self-concept, mediating effects were stronger for the path from cognitive genetics to academic
achievement. This is consistent with observations of moderate associations between cognitive and
non-cognitive skills across compulsory education26,29. In fact, although the term non-cognitive skills
was coined to highlight their independent contribution to academic performance beyond cognitive
skills, the two constructs are related and developmentally intertwined. These results also align with
propositions that non-cognitive skills like interest, curiosity and self-discipline help students leverage
their cognitive competencies more effectively, therefore facilitating academic performance45
Our time-lagged analysis confirmed that the mediating role of non-cognitive skills lasted throughout
compulsory education. These findings reinforce the pivotal role of non-cognitive skills in translating
genetic dispositions into academic achievement, and underscore the importance of fostering non-
cognitive skills to enhance students’ long term academic outcomes, consistent with previous studies
on the effectiveness of interventions for improving non-cognitive skills. For instance, interventions
focusing on self-control and social skills at age 7 were found to have a positive impact on academic
achievement in early adulthood46. A recent systematic review and meta-analysis showed that early-life
interventions aimed at improving self-regulation and perseverance were associated with modest
improvements in later-life academic and psychosocial outcomes47.
Our within-family analyses provided us with a more nuanced understanding of the potential
mechanisms through which genetic dispositions and non-cognitive experiential processes combine
and contribute to differences between students in academic achievement. Examining differences
between siblings, dizygotic twins in this case, we were able to separate the effects of family-wide
processes from those processes that are unique to each child within a family. We found that the
mediating role of non-cognitive skills could be observed not only between-families, but also when
looking at differences between siblings, accounting for a substantial portion of the total effect. Each
children’s unique genetic dispositions were associated with academic achievement partly through
their distinct non-cognitive profiles beyond those genetic, environmental and demographic factors that
are common to family members, which include assortative mating34, population stratification33, and
passive gene-environment correlation. This underscores the key role played by individual-specific
non-cognitive profiles in propelling children down different learning pathways. Our findings are in
line with transactional models of human development rooted in evocative/active gene-environment
correlation36,48 that proposes that, partly in line with their genetic dispositions, children actively seek
out different environmental experiences on the basis of their non-cognitive characteristics (e.g.,
interests, preferences, and aptitudes), and in turn, these will lead to different learning outcomes.
A number of limitations should be acknowledged. First, the reliance on polygenic scores as proxies
for genetic influences on academic achievement is an inherent limitation, as these scores are imperfect
representations of the complex genetic architecture underlying academic performance14,49. This is
particularly relevant to our non-cognitive polygenic score analyses, since this score is indirectly
constructed by isolating those genetic variants related to education that are not associated with
cognitive skills, and not directly from a GWAS of actual non-cognitive measures29. It would be of
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
12
interest to replicate these results when adequately powered GWAS of non-cognitive skills will
become available. Second, our non-cognitive measures may not fully capture the breadth and depth of
factors influencing individuals’ non-cognitive profiles relevant to academic success 23,50. Extant
literature on the links between non-cognitive skills and educational outcomes is complex and
characterised by inconsistencies, low-quality studies and publication bias47. Greater clarity regarding
the taxonomy of non-cognitive skills that relate to education, better measures, and high-quality
longitudinal studies are urgently needed.
Third, we cannot exclude potential genetic confounding in our mediation analyses. Mediation
pathways may be under-corrected for genetic influences in phenotypic variables51–53, which could lead
to an overestimation of the genetic effects mediated by non-cognitive skills. Fourth, adjusting for
heritable covariates such as family SES and general cognitive ability, could introduce collider bias,
potentially distorting the relationship between non-cognitive skills and academic achievement54. Fifth,
our study was conducted within a UK-based twin sample, primarily focused on individuals of White-
European ancestry, which, although representative of the UK population for their birth cohort, limits
the generalisability of our findings to other populations with different socio-cultural and ancestral
backgrounds. Increasingly diverse samples and methodological advances in genetic research, such as
multi-ancestry GWASs55 and novel GWAS methods56, will help bridge this gap in future studies.
Lastly, with the emergence of new methodologies to differentiate between-family and within-family
genetic effects, we plan to extend our research and continue triangulating evidence using multiple
approaches57.
Methods
Participants
Our study involved twins who were part of the Twins Early Development Study (TEDS)58. TEDS has
gathered information from twins born in England and Wales between 1994 and 1996, along with their
parents, at various intervals spanning childhood, adolescence, and early adulthood, commencing from
birth. The initial data collection involved the participation of over 13,000 twin pairs, and after almost
three decades, TEDS maintains an active membership of over 10,000 families. TEDS sample remains
largely representative of the demographic characteristics of the UK population for their respective
generation in terms of ethnicity and socio-economic status59,60. The subsample included in our current
analyses comprised 5,016 individuals (2,895 DZ sibling pairs for the within-family design) whose
families had contributed data on academic achievement, non-cognitive skills, and had available
genotype information. The gender distribution within the sample was approximately 51.1% female
and 48.9% male. Our examination spanned four waves of data collection: age 7 (Mean age = 7.20),
age 9 (Mean age = 9.03), age 12 (Mean age = 11.52), and age 16 (Mean age = 16.31). Mean ages are
derived from teacher assessment of academic achievement. Participants with significant medical,
genetic, or neurodevelopmental conditions were excluded from our analyses, resulting in a variable
sample size ranging from 1,516 to 5,016 due to the inclusion of distinct variables.
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
13
Measures
Data were collected by means of questionnaires and tests administered to parents, teachers, and the
twins themselves by post, telephone, and online, as described in detail in this overview of the TEDS
study60 and in the TEDS data dictionary (https://www.teds.ac.uk/datadictionary/home.htm).
Academic achievement
At ages 7, 9 and 12, academic achievements were provided by the teachers who assessed students’
performance based on the UK National Curriculum guidelines designed by the National Foundation
for Educational Research (NFER; http://www.nfer.ac.uk/index.cfm) and the Qualifications and
Curriculum Authority (QCA; http://www.qca.org.uk). At age 7, academic achievement was measured
as a composite of standardised academic performance across two subjects: English and mathematics.
At age 9 and 12, a composite score of standardised academic performance across three subjects:
English, mathematics, and science.
Academic achievement at age 16 was measured as the mean grade score for the General Certificate of
Secondary Education (GCSE) three subjects (English, mathematics and science). GCSEs are
standardized tests taken at age 16, which in the UK for the TEDS sample was the end of compulsory
education. The exams are graded on a scale ranging from A* to G, with a U grade assigned for
unsuccessful attempts. The grades were coded on a scale from 11 (A*) to 4 (G, the lowest passing
grade), and the mean of the grade obtained from the GCSE core subjects was used as our measure of
academic achievement at age 16. Data on GCSE exam results were collected from parental and self-
reports via questionaries sent over mail or telephone. Our previous research has shown that teacher
ratings and self-reported GCSE grades are valid, reliable, and correlate very strongly with
standardized exam scores taken at specific moments in the educational curriculum (Key Stages)
obtained from the National Pupil Database61.
Non-cognitive skills
Below we provide a brief description of all the measures included in the current study. More details
can be found in the TEDS data dictionary (https://www.teds.ac.uk/datadictionary), which provided
detailed description of each measure and information on the items included in each construct.
Education-specific non-cognitive skills
At age 9 data on education-specific non-cognitive skills were collected from twins themselves, their
parents and teachers. Measures of academic self-perceived ability62 , academic interest62 and the
Classroom Environment Questionnaire (CEQ63) were available from all raters. The CEQ included the
following subscales rated by parents and twins: (1) CEQ classroom satisfaction scale; (2) CEQ
educational opportunities scale; (3) CEQ adventures scales, assessing enjoyment of learning. Ratings
on the CEQ classroom satisfaction scale were also provided by the teachers.
At age 12 data on education-specific non-cognitive skills were collected from self-reports, parents and
teachers. We collected the following measures: academic interest62, academic self-perceived ability62,
the literacy environment questionnaire64,65 and the mathematics environment questionnaire66. The
questionnaires asked several questions related to literacy and mathematics, for example: Reading is
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
14
one of my favourite activities; When I read books, I learn a lot; and In school, how often do you do
maths problems from text books? all rated on a four-point Likert scale.
At age 16 education-specific non-cognitive skills were assessed via self-reports provided by the twins.
The battery of education-specific non-cognitive constructs included the following measures:
Academic self-concept: 10 items of a brief academic self-concept scale (adapted from67, e.g., ‘I like
having difficult work to do and I am clever’) rated on a 5-point Likert scale.
School engagement68: a composite includes 5 subscales: future aspirations and goals; control and
relevance of schoolwork; teacher-student relations; peer support for learning; family support for
learning. The scale includes items rated on four-point Likert scale, such as: ‘School is important for
achieving my future goals’, ‘I feel like I have a say about what happens to me at school’, ‘I enjoy
talking to the teachers at my school’, and ‘When I have problems at school, my family/carer(s) are
willing to help me’).
Grit was measured with 8 items from the Short Grit Scale (GRIT-S)69 asking the twins to report on
their academic perseverance answering questions (e.g., ‘Setbacks don’t discourage me, and I am a
hard worker’) rated on a 5-point Likert scale.
Academic ambition70 was assessed with 5 items asking participants to rate statements like the
following: ‘I am ambitious and achieving something of lasting importance is the highest goal in life’,
on a 5-point Likert scale.
Time spent on mathematics was measured with 3 items asking participants how much time every
week they spent in, including ‘Regular lessons in mathematics at school’, ‘Out-of school-time lessons
in mathematics’, and ‘Study or homework in mathematics by themselves’.
Mathematics self-efficacy (OECD Programme for International Student Assessment (PISA)) was
assessed with 8 items rated on four-point Likert scale. The scale including questions to measure how
confident the student felt about having to conduct different mathematics tasks (e.g., ‘Calculating how
many square metres of tiles you need to cover a floor and Understanding graphs presented in
newspapers’).
Mathematics interest (OECD Programme for International Student Assessment (PISA)) was
assessed with 3 questions, asking participants about how their interest in mathematics, such as: ‘I do
mathematics because I enjoy it and I am interested in the things I learn in mathematics’.
Curiosity was assessed with 7 items71 rated on 7-point Likert scale. Students were asked to rate
statements including: ‘When I am actively interested in something, it takes a great deal to interrupt
me and Everywhere I go’, ‘I am looking out for new things or experiences’.
Attitudes towards school was assessed using the OECD Programme for International Student
Assessment (PISA) attitudes to school measure which included 4 items rated on a 4-point Likert scale,
such as: ‘School has helped give me confidence to make decisions’ and ‘School has taught me things
which could be useful in a job’.
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
15
Self-regulation
The Strengths and Difficulties Questionnaire (SDQ)72 scale was used to evaluate emotional and
behavioural self-regulation across all ages. Data on domain-general self-regulation abilities was
collected from parents, teachers, and self-reports of the twins. The SDQ comprises five subscales:
conduct problems, peer problems, hyperactivity, emotional difficulties, and prosocial behaviour.
Scores from all subscales, except prosocial behaviour, were reversed to ensure higher scores indicate
better self-regulation skills. Parents rated self-regulation skills at age 7, while assessments at ages 9
and 12 involved inputs from parents, teachers, and the twins. At age 16, twins reported their own self-
regulation skills.
Cognitive ability
Cognitive ability at age 7 was evaluated through four tests administered via telephone by trained
research assistants. Verbal cognitive ability was measured using two tests adapted from the Wechsler
Intelligence Scale for Children (WISC-III)73: a 13-item Similarity test and an 18-item Vocabulary test.
Nonverbal cognitive ability was evaluated with a 9-item Conceptual Groupings Test74 and a 21-item
WISC Picture Completion Test73. Composite scores for verbal and nonverbal abilities were calculated
by taking the average of the standardized scores within each respective domain. Furthermore, a
general cognitive ability composite (g) was established by averaging the standardized scores obtained
from the two verbal and two nonverbal tests.
Cognitive ability at age 9 was evaluated through four booklets sent to TEDS families by mail. Verbal
skills were assessed using the initial sections (20 items) of the WISC-III-PI Words and General
Knowledge tests75. Nonverbal abilities were measured with the Shapes test (CAT3 Figure
Classification76 and the Puzzle test (CAT3 Figure Analogies)76. Composite scores for verbal and
nonverbal skills were created by averaging the standardized scores within each area. Additionally, a g
composite was calculated by averaging the standardized scores from the two verbal and two
nonverbal tests.
Cognitive ability at age 12 was evaluated through four tests administered online. Verbal skills were
assessed using the complete versions of the verbal tests conducted at age 9: the entire 30 items from
the WISC-III-PI Words test75 and 30 items from the WISC-III-PI General Knowledge test75.
Nonverbal abilities were measured with the 24-item Pattern test (derived from Raven’s Standard
Progressive Matrices)77 and the 30-item Picture Completion test (WISC-III-UK)73. Composite scores
for verbal and nonverbal abilities were derived by averaging the standardized scores within each
respective domain. Furthermore, a g composite was computed by averaging the standardized scores
from the two verbal and two nonverbal tests.
Cognitive ability at age 16 was assessed through a comprehensive evaluation consisting of one verbal
test and one nonverbal test administered online. Verbal proficiency was measured using a modified
iteration of the Mill Hill Vocabulary test78, while nonverbal abilities were gauged through an adapted
version of the Raven’s Standard Progressive Matrices test. A g composite was determined by
averaging the standardized scores derived from the results of both tests.
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
16
Family socio-economic status
We used socioeconomic status data collected at the first contact for the models at ages 7, 9, and 12.
For the models at age 16, we used the socioeconomic status data available at that age. Specifically, at
first contact, TEDS twins' parents were contacted via postal mail and asked to complete a
questionnaire concerning their educational background, employment status, and the age at which the
mothers had their first child. A composite measure of socioeconomic status was created by
standardizing these three variables and calculating their average. At the age of 16, socioeconomic
status information was collected through a web-based questionnaire, and a comprehensive score was
generated by averaging the standardized values of five items: household income, the highest
educational attainment of both parents, and the employment status of both parents.
Genetic data
Two different genotyping platforms were utilized due to genotyping being conducted in two distinct
phases, with a 5-year interval between them. Affymetrix GeneChip 6.0 SNP arrays were employed to
genotype 3,665 individuals, while Illumina HumanOmniExpressExome-8v1.2 arrays were used for
genotyping 8,122 individuals (including 3,607 DZ co-twin samples). Following quality control
procedures, genotypes from a total of 10,346 samples (comprising 3,320 DZ twin pairs and 7,026
unrelated individuals) were deemed acceptable, with 3,057 individuals genotyped on Affymetrix and
7,289 individuals genotyped on Illumina arrays. The final dataset included 7,363,646 genotyped or
well-imputed SNPs. For further details regarding the handling of these samples, refer to the provided
source79.
Polygenic score
Following the implementation of DNA quality control measures recommended for chip-based
genomic data , we developed genome-wide polygenic scores (PGS) utilizing summary statistics
obtained from four distinct genome-wide association studies: educational attainment14, cognitive
ability29,80 and non-cognitive skills29,31. Each PGS was computed as the weighted sum of an
individual's genotype across all single nucleotide polymorphisms (SNPs), with adjustments made for
linkage disequilibrium using LDpred81. LDpred is a Bayesian shrinkage method that adjusts for local
linkage disequilibrium (the correlation between SNPs) by utilizing data from a reference panel. In our
study, we used the target sample (TEDS), limited to unrelated individuals, along with a prior
reflecting the genetic architecture of the trait. We constructed polygenic scores (PGS) using an
infinitesimal prior, assuming that all SNPs contribute to the genetic architecture of the trait. This
approach has been shown to work well for highly polygenic traits like educational attainment (EA). In
our regression analyses, both the cognitive (Cog) and non-cognitive (NonCog) PGSs were included in
multiple regression models, along with covariates such as age, sex, the first ten principal components
of ancestry, and genotyping chip and batch. To address the non-independence of observations, we
used the generalized estimating equation (GEE) approach82 from the R package.
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
17
Analytic strategies
All scripts will be made available on our research lab GitHub page
(https://github.com/CoDEresearchlab).
Data preparation
To ensure thorough control for potential confounding factors and to effectively evaluate the mediating
impact of these measures on the association between polygenic scores and academic performance, all
non-cognitive skill measures and academic achievements underwent regression against age and sex
effects based on the information on the same measure.
Construction of latent factors of non-cognitive skills and general cognitive ability
Confirmatory factor analysis (CFA)83 was utilised to establish latent constructs representing general
cognitive ability and non-cognitive skills across various developmental stages. In alignment with
established research on general cognitive ability (g) and prior investigations within the TEDS dataset,
we formed a single factor for g at each developmental phase. These g factors were constructed by
aggregating the weighted loadings of two verbal and two nonverbal tests, as outlined in the Measures
section and Supplementary Table 23.
CFA was also utilised to create dimensions reflecting non-cognitive skills. Drawing from previous
meta-analytic research on the non-cognitive factors influencing educational outcomes, we adopted a
theoretical differentiation between education-specific non-cognitive traits (such as academic related
attitudes, motivation, and goals) and broader, more universally applicable measures of self-regulation
(encompassing behavioural and emotional regulation). Consequently, distinct factors were established
for a) education-specific non-cognitive skills (NCS) and b) domain-general self-regulation skills,
considering the various measures available at each age for each evaluator. Please refer to
Supplementary Table 24 and 25 for the factor loadings and model fit indices associated with these
constructs.
Mediation analyses
Following the removal of outliers (scores outside +/- 4 standard deviations), we employed Structural
Equation Modelling (SEM) for conducting mediation analyses84 using lavaan in R85. A
comprehensive description of the mediation models is provided in Supplementary Note 2. These
mediation models enabled us to dissect the effect size of each polygenic score's prediction on
academic achievement throughout development into direct and indirect effects mediated by both
latent dimensions of, and individual non-cognitive skill measures. To address the risk of multiple
testing, we employed the Benjamini-Hochberg false discovery rate correction (FDR)86. Furthermore,
to address the non-independence of observations in the sample (i.e., relatedness), we randomly
selected one twin from each pair for analysis. Also, we applied bootstrap method87 within lavaan
package, which allows us to assess the variability of our estimates without relying on the assumption
of normality in the sampling distribution.
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
18
Cross-sectional analyses
We first conducted mediation models for the non-cognitive skills and academic achievement at the
same developmental stage across compulsory education, examining the pathway from polygenic
scores to non-cognitive skills and academic achievement measured at the same age. For example, we
investigated whether the education-specific non-cognitive skills assessed at age 9 mediated the
polygenic score effects on academic achievement at the same age. We selected mediators based on
data availability during the same collection wave as each academic achievement outcome.
Longitudinal mediation analyses
We further applied longitudinal models, using non-cognitive skills measured at earlier ages as
mediators for later academic achievement to investigate whether the non-cognitive skills mediated
genetic disposition with academic achievement longitudinally. For instance, we conducted mediation
models by including education-specific non-cognitive skills assessed at age 9 and academic
achievement at ages 12 and 16.
Two-mediator mediation analyses
To account for potential confounding effects, we utilized two-mediator mediation models88 to expand
our examination of mediation effects and explore whether the mediating role of non-cognitive skills
was primarily influenced by either family socioeconomic status (SES) or general cognitive ability (g).
Therefore, we reiterated our mediation models by simultaneously considering each non-cognitive skill
alongside family SES and the g composite.
Multi-level mediation analysis: separating between from within-family effects
We further utilized 1-1-1 two-level mediation models89 to disentangle within-family effects from
between-family effects. This statistical framework enabled us to assess the indirect impact of a
predictor on an outcome while considering clustered mediation data by mean-centring the siblings’
measurements and taking the departure from the mean for each sibling. In our analyses, we organized
our data by family, with each family representing a cluster comprising two members (the two
dizygotic twins). By employing 1-1-1 multilevel mediation models, we could separate the between-
sibling polygenic score effects, as well as mediation effects. Between-family effects estimated from
samples of unrelated individuals are typically biased by sibling-invariant family level demographic
factors such as genetic nurture and population stratification and passive gene-environment correlation
17,19,90. Within-family analyses rely on the randomization of allele transmission during meiosis,
ensuring that siblings have an equal chance of inheriting any given allele, independent of
environmental processes. Consequently, genetic differences between siblings are unaffected by
environmental factors making within-family effects robust to major sources of environmental
contributions. Our within-family mediation estimates are therefore indicative of how each sibling’s
genetic disposition towards education combines with individual-specific socio-emotional processes to
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
19
lead to differential academic outcomes through evocative/active gene-environment correlation
processes.
In these analyses, only dizygotic (DZ) twins were included, as our objective was to explore how
differences in polygenic scores between siblings predicted variation in academic achievement through
disparities in the non-cognitive skills.
References
1. Cutler, D. M. & Lleras-Muney, A. Education and Health: Insights from International
Comparisons. Working Paper at https://doi.org/10.3386/w17738 (2012).
2. Murray-Harvey, R. Relationship influences on students’ academic achievement, psychological
health and well-being at school. Educ. Child Psychol. 27, 104–115 (2010).
3. Watts, T. W. Academic Achievement and Economic Attainment: Reexamining Associations
Between Test Scores and Long-Run Earnings. AERA Open 6, 2332858420928985 (2020).
4. Currie, J. & Thomas, D. Early Test Scores, Socioeconomic Status and Future Outcomes. Working
Paper at https://doi.org/10.3386/w6943 (1999).
5. Montez, J. K. & Hayward, M. D. Cumulative childhood adversity, educational attainment, and
active life expectancy among U.S. adults. Demography 51, 413–435 (2014).
6. Malanchini, M., Rimfeld, K., Allegrini, A. G., Ritchie, S. J. & Plomin, R. Cognitive ability and
education: How behavioural genetic research has advanced our knowledge and understanding of
their association. Neurosci. Biobehav. Rev. 111, 229–245 (2020).
7. Rimfeld, K. et al. The stability of educational achievement across school years is largely
explained by genetic factors. Npj Sci. Learn. 3, 1–10 (2018).
8. Branigan, A. R., McCallum, K. J. & Freese, J. Variation in the Heritability of Educational
Attainment: An International Meta-Analysis. Soc. Forces 92, 109–140 (2013).
9. Trzaskowski, M. et al. DNA Evidence for Strong Genome-Wide Pleiotropy of Cognitive and
Learning Abilities. Behav. Genet. 43, 267–273 (2013).
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
20
10. Krapohl, E. & Plomin, R. Genetic link between family socioeconomic status and children’s
educational achievement estimated from genome-wide SNPs. Mol. Psychiatry 21, 437–443
(2016).
11. Dudbridge, F. Power and Predictive Accuracy of Polygenic Risk Scores. PLOS Genet. 9,
e1003348 (2013).
12. Belsky, D. W. & Harden, K. P. Phenotypic Annotation: Using Polygenic Scores to Translate
Discoveries From Genome-Wide Association Studies From the Top Down. Curr. Dir. Psychol.
Sci. 28, 82–90 (2019).
13. Okbay, A. et al. Polygenic prediction of educational attainment within and between families from
genome-wide association analyses in 3 million individuals. Nat. Genet. 1–13 (2022)
doi:10.1038/s41588-022-01016-z.
14. Lee, J. J. et al. Gene discovery and polygenic prediction from a genome-wide association study of
educational attainment in 1.1 million individuals. Nat. Genet. 50, 1112–1121 (2018).
15. Allegrini, A. G. et al. Genomic prediction of cognitive traits in childhood and adolescence. Mol.
Psychiatry 24, 819–827 (2019).
16. von Stumm, S. et al. Predicting educational achievement from genomic measures and
socioeconomic status. Dev. Sci. 23, e12925 (2020).
17. Zhou, Q. et al. Gene-environment correlation: The role of family environment in academic
development. 2024.01.05.574339 Preprint at https://doi.org/10.1101/2024.01.05.574339 (2024).
18. Plomin, R., DeFries, J. C. & Loehlin, J. C. Genotype-environment interaction and correlation in
the analysis of human behavior. Psychol. Bull. 84, 309–322 (1977).
19. Kong, A. et al. The nature of nurture: Effects of parental genotypes. Science 359, 424–428 (2018).
20. Koellinger, P. D. & Harden, K. P. Using nature to understand nurture. Science 359, 386–387
(2018).
21. Wertz, J. et al. Using DNA From Mothers and Children to Study Parental Investment in
Children’s Educational Attainment. Child Dev. 91, 1745–1761 (2020).
22. Heckman, J. J. & Rubinstein, Y. The Importance of Noncognitive Skills: Lessons from the GED
Testing Program. Am. Econ. Rev. 91, 145–149 (2001).
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
21
23. Schneider, M. & Preckel, F. Variables associated with achievement in higher education: A
systematic review of meta-analyses. Psychol. Bull. 143, 565–600 (2017).
24. Krapohl, E. et al. The high heritability of educational achievement reflects many genetically
influenced traits, not just intelligence. Proc. Natl. Acad. Sci. 111, 15273–15278 (2014).
25. Chamorro-Premuzic, T., Harlaar, N., Greven, C. U. & Plomin, R. More than just IQ: A
longitudinal examination of self-perceived abilities as predictors of academic performance in a
large sample of UK twins. Intelligence 38, 385–392 (2010).
26. Malanchini, M., Engelhardt, L. E., Grotzinger, A. D., Harden, K. P. & Tucker-Drob, E. M. ‘Same
but different’: Associations between multiple aspects of self-regulation, cognition and academic
abilities. J. Pers. Soc. Psychol. 117, 1164–1188 (2019).
27. Muenks, K., Wigfield, A., Yang, J. & O’Neal, C. How True Is Grit? Assessing Its Relations to
High School and College Students’ Personality Characteristics, Self-Regulation, Engagement,
and Achievement. J. Educ. Psychol. 109, 599–620 (2017).
28. Tangney, J. P., Baumeister, R. F. & Boone, A. L. High self-control predicts good adjustment, less
pathology, better grades, and interpersonal success. J. Pers. 72, 271–324 (2004).
29. Malanchini, M. et al. Genetic associations between non-cognitive skills and academic
achievement over development. Nat. Hum. Behav. 8, 2034–2046 (2024).
30. Belsky, D. W. et al. The Genetics of Success: How Single-Nucleotide Polymorphisms Associated
With Educational Attainment Relate to Life-Course Development. Psychol. Sci. 27, 957–972
(2016).
31. Demange, P. A. et al. Investigating the genetic architecture of noncognitive skills using GWAS-
by-subtraction. Nat. Genet. 53, 35–44 (2021).
32. Price, T. S. & Jaffee, S. R. Effects of the family environment: Gene-environment interaction and
passive gene-environment correlation. Dev. Psychol. 44, 305–315 (2008).
33. Baier, T. & Lang, V. The Social Stratification of Environmental and Genetic Influences on
Education: New Evidence Using a Register-Based Twin Sample. Sociol. Sci. 6, 143–171 (2019).
34. Robinson, M. R. et al. Genetic evidence of assortative mating in humans. Nat. Hum. Behav. 1, 1–
13 (2017).
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
22
35. Selzam, S. et al. Comparing Within- and Between-Family Polygenic Score Prediction. Am. J.
Hum. Genet. 105, 351–363 (2019).
36. Tucker-Drob, E. M. & Harden, K. P. A behavioral genetic perspective on noncognitive factors
and academic achievement. Genet. Ethics Educ. 134–158 (2017).
37.
İ
nan-Kaya, G. & Rubie-Davies, C. M. Teacher classroom interactions and behaviours: Indications
of bias. Learn. Instr. 78, 101516 (2022).
38. Chang, C.-C., Tseng, K.-H. & Lou, S.-J. A comparative analysis of the consistency and difference
among teacher-assessment, student self-assessment and peer-assessment in a Web-based portfolio
assessment environment for high school students. Comput. Educ. 58, 303–320 (2012).
39. Flake, J. K. & Petway II, K. T. Methodologies for Investigating and Interpreting Student–Teacher
Rating Incongruence in Noncognitive Assessment. Educ. Meas. Issues Pract. 38, 63–77 (2019).
40. Marsh, H., Trautwein, U., Ludtke, O., Koller, O. & Baumert, J. Academic self-concept, interest,
grades, and standardized test scores: Reciprocal effects models of causal ordering. CHILD Dev.
76, 397–416 (2005).
41. von Stumm, S., Hell, B. & Chamorro-Premuzic, T. The Hungry Mind: Intellectual Curiosity Is
the Third Pillar of Academic Performance. Perspect. Psychol. Sci. 6, 574–588 (2011).
42. Lazonder, A. W. & Harmsen, R. Meta-Analysis of Inquiry-Based Learning: Effects of Guidance.
Rev. Educ. Res. 86, 681–718 (2016).
43. Hattie, J. & Timperley, H. The Power of Feedback. Rev. Educ. Res. 77, 81–112 (2007).
44. Zimmerman, B. J. Goal setting: A key proactive source of academic self-regulation. in Motivation
and self-regulated learning: Theory, research, and applications 267–295 (Lawrence Erlbaum
Associates Publishers, Mahwah, NJ, US, 2008).
45. Heckman, J. J. & Kautz, T. Fostering and Measuring Skills: Interventions That Improve
Character and Cognition. Working Paper at https://doi.org/10.3386/w19656 (2013).
46. Algan, Y., Beasley, E., Vitaro, F. & Tremblay, R. The Impact of Non-Cognitive Skills Training
on Academic and Non-academic Trajectories: From Childhood to Early Adulthood. (2014).
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
23
47. Smithers, L. G. et al. A systematic review and meta-analysis of effects of early life non-cognitive
skills on academic, psychosocial, cognitive and health outcomes. Nat. Hum. Behav. 2, 867–880
(2018).
48. Plomin, R. Genotype-Environment Correlation in the Era of DNA. Behav. Genet. 44, 629–638
(2014).
49. Non, A. L. & Cerdeña, J. P. Considerations, Caveats, and Suggestions for the Use of Polygenic
Scores for Social and Behavioral Traits. Behav. Genet. 54, 34–41 (2024).
50. Gutman, L. M. & Schoon, I. The Impact of Non-Cognitive Skills on Outcomes for Young People.
A Literature Review. Education Endowment Foundation: London, UK.
https://educationendowmentfoundation.org.uk/evidence-summaries/evidence-reviews/essential-
life-skills/ (2013).
51. Pingault, J.-B. et al. Research Review: How to interpret associations between polygenic scores,
environmental risks, and phenotypes. J. Child Psychol. Psychiatry 63, 1125–1139 (2022).
52. Allegrini, A. G. et al. Multivariable G-E interplay in the prediction of educational achievement.
PLOS Genet. 16, e1009153 (2020).
53. MacKinnon, D. P., Krull, J. L. & Lockwood, C. M. Equivalence of the mediation, confounding
and suppression effect. Prev. Sci. Off. J. Soc. Prev. Res. 1, 173–181 (2000).
54. Akimova, E. T., Breen, R., Brazel, D. M. & Mills, M. C. Gene-environment dependencies lead to
collider bias in models with polygenic scores. Sci. Rep. 11, 9457 (2021).
55. Yengo, L. et al. A saturated map of common genetic variants associated with human height.
Nature 610, 704–712 (2022).
56. Campos, A. I. et al. Boosting the power of genome-wide association studies within and across
ancestries by using polygenic scores. Nat. Genet. 1–8 (2023) doi:10.1038/s41588-023-01500-0.
57. Howe, L. J. et al. Within-sibship genome-wide association analyses decrease bias in estimates of
direct genetic effects. Nat. Genet. 54, 581–592 (2022).
58. Lockhart, C. et al. Twins Early Development Study (TEDS): A genetically sensitive investigation
of mental health outcomes in the mid-twenties. JCPP Adv. 3, e12154 (2023).
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
24
59. Lockhart, C. et al. Twins Early Development Study (TEDS): A genetically sensitive investigation
of mental health outcomes in the mid-twenties. JCPP Adv. 3, e12154 (2023).
60. Rimfeld, K. et al. Twins Early Development Study: A Genetically Sensitive Investigation into
Behavioral and Cognitive Development from Infancy to Emerging Adulthood. Twin Res. Hum.
Genet. 22, 508–513 (2019).
61. Rimfeld, K. et al. Teacher assessments during compulsory education are as reliable, stable and
heritable as standardized test scores. J. Child Psychol. Psychiatry 60, 1278–1288 (2019).
62. Spinath, B., Spinath, F. M., Harlaar, N. & Plomin, R. Predicting school achievement from general
cognitive ability, self-perceived ability, and intrinsic value. Intelligence 34, 363–374 (2006).
63. Walker, S. O. & Plomin, R. Nature, Nurture, and Perceptions of the Classroom Environment as
They Relate to Teacher
Assessed Academic Achievement: A twin study of nine
year
olds.
Educ. Psychol. 26, 541–561 (2006).
64. Perie, M., Grigg, W. & Donahue, P. The Nation’s Report Card: Reading 2005. (2005).
65. The Nation’s Report Card: Reading 2005.
https://nces.ed.gov/pubsearch/pubsinfo.asp?pubid=2006451 (2005).
66. Perie, M., Grigg, W. & Dion, G. The Nation’s Report Card: Mathematics 2005. (2005).
67. Burden, R. Assessing Children’s Perceptions of Themselves as Learners and Problem-Solvers:
The Construction of the Myself-as-Learner Scale (MALS). Sch. Psychol. Int. 19, 291–305 (1998).
68. Appleton, J. J., Christenson, S. L., Kim, D. & Reschly, A. L. Measuring cognitive and
psychological engagement: Validation of the Student Engagement Instrument. J. Sch. Psychol. 44,
427–445 (2006).
69. Duckworth, A. L. & Quinn, P. D. Development and Validation of the Short Grit Scale (Grit–S). J.
Pers. Assess. 91, 166–174 (2009).
70. Duckworth, A. L., Peterson, C., Matthews, M. D. & Kelly, D. R. Grit: Perseverance and passion
for long-term goals. J. Pers. Soc. Psychol. 92, 1087–1101 (2007).
71. Kashdan, T. B., Rose, P. & Fincham, F. D. Curiosity and Exploration: Facilitating Positive
Subjective Experiences and Personal Growth Opportunities. J. Pers. Assess. 82, 291–305 (2004).
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
25
72. Goodman, R. The Strengths and Difficulties Questionnaire: A Research Note. J. Child Psychol.
Psychiatry 38, 581–586 (1997).
73. Wechsler, D. Wechsler Intelligence Scale for Children; Manual. v, 113 (The Psychological Corp.,
Oxford, England, 1949).
74. McCarthy, D. Manual for the McCarthy Scales of Children’s Abilities. (The Psychological
Corporation, New York, New York, 1972).
75. Kaplan, E. Wechsler Intelligence Scale for Children as a Process Instrument: -WISC-III PI.
(Psychological Corporation, 1999).
76. Smith, P., Fernandes, C. & Strand, S. Cognitive abilities test 3 (CAT3). Windsor NferNELSON
(2001).
77. Raven, J. C., Court, J. H. & Raven, J. Manual for Raven’s Progressive Matrices and Vocabulary
Scales. Section 3, Standard Progressive Matrices. (Oxford Psychologists Press, Oxford, 1996).
78. Raven, J. C., Raven, J. E. & Court, J. H. Mill Hill Vocabulary Scale. (Psychological Corporation,
1989).
79. Selzam, S. et al. Evidence for gene-environment correlation in child feeding: Links between
common genetic variation for BMI in children and parental feeding practices. PLOS Genet. 14,
e1007757 (2018).
80. Savage, J. E. et al. Genome-wide association meta-analysis in 269,867 individuals identifies new
genetic and functional links to intelligence. Nat. Genet. 50, 912–919 (2018).
81. Vilhjálmsson, B. J. et al. Modeling Linkage Disequilibrium Increases Accuracy of Polygenic
Risk Scores. Am. J. Hum. Genet. 97, 576–592 (2015).
82. Vanegas, L. H., Rondón, L. M. & Paula, G. A. Generalized Estimating Equations using the new R
package glmtoolbox. R J. 15, 105–133 (2023).
83. Brown, T. A. Confirmatory Factor Analysis for Applied Research, Second Edition. (Guilford
Publications, 2015).
84. Gunzler, D., Chen, T., Wu, P. & Zhang, H. Introduction to mediation analysis with structural
equation modeling. Shanghai Arch. Psychiatry 25, 390–394 (2013).
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
26
85. Rosseel, Y. lavaan: An R Package for Structural Equation Modeling. J. Stat. Softw. 48, 1–36
(2012).
86. Benjamini, Y. & Hochberg, Y. On the Adaptive Control of the False Discovery Rate in Multiple
Testing With Independent Statistics. J. Educ. Behav. Stat. 25, 60–83 (2000).
87. DiCiccio, T. J. & Efron, B. Bootstrap confidence intervals. Stat. Sci. 11, 189–228 (1996).
88. VanderWeele, T. J. & Vansteelandt, S. Mediation Analysis with Multiple Mediators. Epidemiol.
Methods 2, 95–115 (2014).
89. Tofighi, D., West, S. G. & MacKinnon, D. P. Multilevel mediation analysis: The effects of
omitted variables in the 1–1–1 model. Br. J. Math. Stat. Psychol. 66, 290–307 (2013).
90. Morris, T. T., Davies, N. M., Hemani, G. & Smith, G. D. Population phenomena inflate genetic
associations of complex social traits. Sci. Adv. 6, eaay0328 (2020).
Code availability statement
The code is available at: https://github.com/CoDEresearchlab
Author information
Corresponding authors: Quan Zhou & Margherita Malanchini
School of Biological and Behavioural Sciences, Queen Mary University of London
Quan Zhou, Margherita Malanchini
Social, Genetic and Developmental Psychiatry Centre, King’s College London
Andrea G. Allegrini, Kaili Rimfeld, Robert Plomin, Margherita Malanchini
Division of Psychology and Language Sciences, University College London
Andrea G. Allegrini
Department of Psychology, Royal Holloway University of London
Kaili Rimfeld
School of Philosophy, Psychology and Language Sciences, The University of Edinburgh,
Edinburgh, UK
Jasmin Wertz
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
27
Centre for Longitudinal Studies, Social Research Institute, University College London, London,
UK
Tim Morris
Max Planck Research Group Biosocial – Biology, Social Disparities, and Development; Max
Planck Center for Human Development, Berlin, Germany
Laurel Raffington
Acknowledgements
This study was funded by a starting grant awarded to MM by the School of Biological and
Behavioural Sciences at Queen Mary, University of London. M.M. is supported by a Jacobs
Foundation Research Fellowship. QZ is funded by the QMUL-Chinese Scholarship Council joint PhD
Scholarship. TEDS is supported by a programme grant to RP from the UK Medical Research Council
(MR/V012878/1 and previously MR/M021475/1), with additional support from the US National
Institutes of Health (AG046938).
Author contributions
Conceived and designed the study: M.M., Q.Z. Analysed the data: Q.Z. Supervised the work: M.M..
Wrote the paper: Q.Z., M.M., with helpful contributions from A.A., J.W., T.T.M., L.R., K.R., and
R.P.. All authors contributed to the interpretation of data, provided feedback on drafts, and approved
the final draft.
Competing interests
The authors declare no competing interests.
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.27.640510doi: bioRxiv preprint
