The genetic variations that relate to how children think, learn, and solve problems shift in their associations between early childhood and adolescence, according to new research published in Nature Human Behaviour. By tracking thousands of British children from age 4 to age 16, researchers found that the link between common genetic variants tied to education and test scores grows stronger with age, particularly among high-scoring youths.
In contrast, the negative association between rare, gene-disrupting mutations and intelligence test scores is strongest in early childhood and weakens by adolescence, a pattern that may help explain why some adults carry mutations linked to childhood developmental disorders without showing noticeable impairments themselves.
Common variants, rare mutations, and intelligence
Cognitive ability refers to a person’s general capacity to reason, solve problems, think abstractly, and learn new information, which psychologists typically estimate using standardized intelligence quotient, or IQ, tests. This study focused on two broad classes of variations in human DNA that relate to cognitive differences alongside environmental circumstances. Common genetic variants are frequent single-letter DNA differences found in more than 1 percent of the population that each have a tiny association with traits, whereas rare damaging variants are uncommon mutations that disrupt how a gene builds a specific protein.
Twin and family studies indicate that the overall genetic contribution to intelligence is not fixed across the lifespan. For instance, the authors of a 2014 review noted that the share of intelligence differences explained by genetics rises from roughly 20 percent in infancy to about 60 percent or more in adulthood. Molecular studies have since identified specific genetic regions involved: a 2018 genetic study of 1.1 million adults pinpointed over a thousand common DNA variants tied to years of schooling, allowing scientists to combine millions of tiny genetic effects into a single summary measure known as a polygenic score that predicts between 7 and 10 percent of the variation in cognitive test performance.
Don’t miss the next big psychology finding
Every morning, PsyPost sends a short briefing on the most interesting new peer-reviewed research in psychology and neuroscience.
These genetic factors also intersect with neurodevelopmental conditions, which are disorders affecting early brain growth that can cause developmental delays, learning difficulties, or intellectual disability. For example, an earlier study of people carrying rare deletions on chromosome 16 linked to autism found specific deficits in reading fluency, speech sound processing, fine motor skills, and impulse control even after accounting for lower overall IQ. Alongside rare mutations in single genes, a 2024 study by several of the same researchers found that common genetic variants explain about 10 percent of the variation in risk for rare neurodevelopmental conditions.
Why some parents carry mutations without apparent effects
A persistent puzzle in genetics is why many rare damaging mutations show what scientists call incomplete penetrance, meaning that a child with a developmental disorder often inherits the causal mutation from a parent who appears unaffected. One possibility is that these rare mutations have time-varying effects that impair cognitive function more in early childhood than in adulthood, or that they widen the spread of cognitive outcomes so that only some carriers cross the threshold for a diagnosis.
The research, led by Daniel S. Malawsky and Hilary C. Martin of the Wellcome Sanger Institute in the UK, tracked both common and rare genetic variants side by side across childhood to see how their associations with cognitive performance evolve over time.
Tracking IQ from age 4 to 16
To investigate these developmental trajectories, the researchers analyzed data from three British cohorts: the Avon Longitudinal Study of Parents and Children, the UK Millennium Cohort Study, and the UK Biobank. Their primary longitudinal analysis focused on 6,495 unrelated children of European genetic ancestry from the Avon cohort, which recruited pregnant mothers in the Bristol area in the early 1990s. These children were given age-appropriate Wechsler IQ tests at clinic visits at ages 4, 8, and 16, though only a subset of the cohort was invited to the age-4 test. Among the 6,495 children analyzed, 701 had a measured score at age 4, 5,134 at age 8, and 3,774 at age 16, while 3,895 also had linked national school exam scores in English, math, and science at ages 11 and 14.
Because not every child attended all three clinic visits, the scientists used a statistical procedure called imputation to estimate missing IQ scores across the three ages for all 8,207 children in the broader cohort who had at least one IQ measurement, drawing on related developmental, communication, and demographic records. At age 4, roughly 9 in 10 scores in the main analysis (5,794 of 6,495) were statistical estimates, and in a check where known scores were hidden and re-estimated, the estimates matched the real ones only moderately well, with a median correlation of 0.61.
The researchers ran their analyses both before and after filling in those missing test scores to check whether participant dropout distorted the findings. The team also examined a composite cognitive score from tests given between ages 3 and 7 to up to 5,920 children in the Millennium Cohort Study, alongside fluid intelligence scores from 101,232 adults in the UK Biobank.
Measuring common and rare genetic variation
To measure common genetic variation, the authors calculated three polygenic scores for each participant: one for overall educational attainment, one for its cognitive component, and one for its non-cognitive component, which captures traits like personality and socioeconomic factors associated with staying in school. To evaluate rare genetic variants, the team generated new exome sequencing data, a technique that reads the protein-coding sections of DNA, for 16,103 children and 10,140 parents across the two birth cohorts.
Because individual rare mutations are too scarce to measure on their own, the researchers calculated each person’s rare variant burden by summing up uncommon mutations present in less than 0.1 percent of the population. Each affected gene was weighted by how strongly natural selection has weeded out mutations in that gene over human history.
The researchers created separate rare variant scores for mutations predicted to completely disable a gene, mutations that alter a protein’s structure in a damaging way, and harmless mutations that do not change the resulting protein, which served as a control. In all models, the authors controlled for the child’s sex and broad patterns of genetic ancestry. In a subset of 4,968 parent-child trios with common variant data and 958 trios with rare variant exome data in the Avon cohort, they also controlled for both parents’ genetic scores. This approach separates the direct biological association of the variants a child inherited from indirect family influences, often called genetic nurture, where parents’ genes shape the home environment.
Links to common variants grew stronger with age
Across the Avon cohort, the researchers estimated that the overall share of IQ variation explained by common DNA variants rose from 46 percent at age 4 to 56 percent at age 16, although that increase was not statistically significant. Focusing on the specific summary scores, the positive association between children’s IQ scores and their polygenic scores for educational attainment and cognitive skills grew stronger between ages 4 and 16. The polygenic scores explained about 11 percent of the variation in IQ at age 4, rising to 14 percent by age 16, and a similar upward trend appeared in school exam grades between ages 11 and 14.
In the family trio models, the baseline associations for the educational and cognitive polygenic scores shrank substantially when controlling for parental genetics, and the parents’ own scores were linked to the children’s IQ, indicating that part of the overall association reflects family-level factors. Even so, the age-related increase in the educational and cognitive scores remained intact, which was consistent with direct genetic associations in the children rather than family influences. By contrast, the direct association of the non-cognitive polygenic score with IQ was not statistically significant once parental genetics were accounted for.
Links to rare mutations weakened with age
The pattern for rare damaging variants moved in the opposite direction. Carrying a higher burden of rare gene-disabling or protein-altering mutations was associated with lower IQ scores and lower school grades, with gene-disabling mutations showing an unstandardized negative association 3.2 times larger than protein-altering mutations, while the harmless control mutations showed no link to IQ. However, this negative association weakened between ages 4 and 16, and the weakening with age also appeared in the family trio analysis.
When the team separated rare gene-disabling mutations in evolutionarily important genes into those inherited from a parent and those arising newly in the egg or sperm, inherited variants explained 3.5 times more variation in childhood IQ. Inherited variants were also far more common: 12 percent of children carried one, compared with 1.8 percent for new mutations. The authors note that their sample likely overstated the share from new mutations, and that the pattern contrasts with clinical cohorts of children with developmental disorders, where new mutations play a larger role.
Different patterns at the top and bottom of the IQ range
Next, the scientists used a statistical technique called quantile regression to test whether these genetic scores had different associations at the bottom 5th percentile of the IQ distribution (roughly an IQ of 75), the 50th percentile (average IQ), and the top 95th percentile (roughly an IQ of 125). They found that the growing association of the educational and cognitive polygenic scores with age was concentrated in the upper half of the IQ distribution, whereas the association at the bottom 5th percentile did not show a statistically significant change with age.
Consistent with this childhood trend, data from UK Biobank adults indicated that the direct association of the educational polygenic score was 1.62 times stronger at the 95th percentile of adult intelligence scores than at the 5th percentile.
For rare gene-disabling variants, the negative associations were heavily concentrated among the lowest-scoring children in early life before evening out by adolescence. At age 4, the negative association of rare gene-disabling mutations was 3.9 times greater at the 5th percentile of IQ than at the 95th percentile and 2.2 times greater than at the median, meaning that young children with a high load of rare mutations had a wider spread of cognitive scores, with more of them falling into the lowest range.
By age 16, however, the differences in association strength across the 5th, 50th, and 95th percentiles were no longer statistically significant. Even so, most high-burden children did not have severe impairments: among children in the highest 5 percent of rare variant burden, about 7 percent had an IQ below 70 at age 8, compared with roughly 3 percent of other children.
How genes compare with parental education and birth factors
In their final analysis, the authors compared these genetic measures against three environmental and birth factors: parents’ completed years of education, maternal health complications during pregnancy (such as diabetes, anemia, or preeclampsia), and premature birth (weeks born before 40 weeks of gestation). Together, the genetic measures, parental education, and maternal pregnancy illness explained 22.8 percent of the variation in IQ at age 4 and 25.6 percent at age 16, in a model that left out premature birth because gestational age was available for far fewer children. The direct association of a child’s educational polygenic score with IQ was similar in size to the association with their mother’s or father’s actual educational level.
Rare damaging variants explained 1.40 percent of IQ variation at age 4, four to eight times more than premature birth (0.35 percent) or maternal pregnancy illness (0.18 percent). Those differences were statistically significant, though the authors describe the contributions as broadly comparable. The associations for both birth factors also faded with age.
These findings align with an earlier study linking educational attainment polygenic scores to brain size, which found that people with higher scores tended to have higher cognitive test performance and slightly larger brain volumes, though that work did not track how the association changes across childhood. The results are also consistent with an earlier study of adults with ADHD that linked rare damaging mutations in three genes to lower cognitive performance and reduced educational attainment, although that study focused on specific rare risk genes in adults with ADHD rather than evaluating general rare variant burden across childhood development.
What the findings can and cannot show
As with all research, there are some caveats to consider. Polygenic scores and rare variant burden scores describe statistical trends across thousands of people and cannot predict the cognitive development of an individual child, as children with identical genetic scores still display a wide range of IQ outcomes. Because this study examined a relatively healthy general population, the authors also stress that the weakening pattern for rare variants cannot be assumed to apply to children with diagnosed neurodevelopmental conditions.
In addition, the analyses were restricted to British participants of European genetic ancestry to prevent statistical confounding from population differences and because existing polygenic scores were developed in European samples. Consequently, the results provide no basis for comparing different racial or ethnic groups and may not generalize to other populations. The authors also caution that common- and rare-variant scores are built differently, so strong conclusions about their contrasting patterns are not warranted. Because the rare variant scores were restricted to the 1 percent of the genome that codes for proteins, they miss rare mutations in the rest of the genome, and the observational design means the findings represent statistical associations rather than definitive proof of biological mechanisms.
The longitudinal IQ findings relied on three testing ages from a single birth cohort, and IQ tests in early life are less reliable and less accurate at the extremes of the distribution, although the authors argue that test reliability alone is unlikely to explain the results because the two genetic scores moved in opposite directions. Children with lower educational polygenic scores tended to complete fewer IQ tests, and imputation of missing scores altered some of the percentile-specific polygenic score patterns at age 4 compared to the raw, pre-imputation data. Sensitivity simulations, however, suggested that participant dropout was unlikely to have created the opposing age trends.
“Confidence in the results is somewhat limited by the analyses involving substantial imputation related to missing data, which in some cases altered the results, leaving me with some uncertainty about the interpretation,” Liana Machado, a professor in the Department of Psychology and the Brain Health Research Centre at the University of Otago, who was not involved in the study, told PsyPost. “However, the authors did report pre-imputation results in the supplementary file. Rare variant burden results were further limited by smaller sample sizes. However, the thorough analysis approaches boosted my confidence in the reported results.”
Why the link to rare mutations might weaken over time
The authors say these questions should be revisited once much larger whole-genome sequenced cohorts are available, alongside clinical cohorts that track children diagnosed with neurodevelopmental conditions and their parents over time. They suggest that early cognitive delays in a subset of children carrying rare variants could stem from accompanying physical health challenges, such as congenital heart conditions or vision and hearing impairments, that improve with medical care and targeted interventions, or from biological mechanisms in the developing brain that compensate for a disrupted gene. If early health challenges explain the weakening link with age, the authors predict that this weakening should be strongest in children from less deprived households.
“In terms of what to test next, one thing that stuck out in my mind is the possibility that resolution with age of the link between rare variant burden and lower cognitive scores could be driven by access to health care, which would certainly be worthwhile to investigate as confirmation would highlight a pathway to better outcomes via targeted interventions,” Machado said.
“Alternatively, to the extent that natural compensatory mechanisms kicking in during development explain the attenuating link between rare variant burden and cognition with age, gaining a better understanding of these compensatory mechanisms could provide insight towards potentially fruitful therapeutic pathways that could benefit those at the lower end. Similarly, gaining a better understanding through future research why those with lower cognitive scores do not show a strengthening relationship between genetic score and cognitive performance as they develop could shed light on potentially targetable intervention pathways.”
The study, “Common and rare genetic variant associations with cognitive performance across development in British birth cohorts,” was authored by Daniel S. Malawsky, Mahmoud Koko, Petr Danacek, Wei Huang, Olivia Wootton, Qin Qin Huang, Emma E. Wade, Sarah J. Lindsay, Rosalind Arden, Matthew E. Hurles, and Hilary C. Martin.
