How well can genetics predict intelligence? A large-scale study published in Intelligence explores the power of polygenic scores to predict intelligence, finding that these genetic estimates moderately correlate with IQ. However, the study also highlights critical gaps, including variability across studies and challenges in pinpointing specific genetic influences.
Intelligence, often defined as the ability to learn, reason, and solve problems, is a key factor in predicting various life outcomes, including educational attainment, career success, and overall health. Twin and family studies have consistently shown that genetic differences account for about half of the variance in intelligence among individuals, with genetic influences becoming stronger across the lifespan. However, identifying the specific genetic mechanisms underlying these differences has proven to be a significant challenge.
The advent of genome-wide association studies has opened new doors for exploring the genetic architecture of intelligence. These studies analyze genetic data from large populations to identify DNA variants associated with specific traits, such as intelligence. By aggregating these variants into polygenic scores, researchers can estimate an individual’s genetic propensity for a particular trait.
Polygenic scores have been used to predict various characteristics, but their utility in predicting intelligence remains a topic of considerable debate. The new study aimed to address key questions about the power, reliability, and limitations of polygenic scores in the context of intelligence.
“Intelligence has a real bearing on people’s life outcomes. Understanding what drives people’s differences in intelligence is an important first step in identifying ways to work towards a fairer society,” explained study author Florence Oxley, a postdoctoral research associate at the University of York who is affiliated with the Hungry Mind Lab and the SENFM research project.
“Research has shown that differences in intelligence are partly genetic and partly shaped by environmental experiences. People’s genetic tendency towards a given trait can be captured by so-called polygenic scores, which add together the DNA variants a person possesses that are linked to that trait. Polygenic scores for intelligence can predict how well someone will score on an IQ test. In our research, we explored how powerful and accurate polygenic score predictions of intelligence are across studies.”
The researchers conducted their study using a meta-analytic approach, pooling data from multiple independent studies to evaluate the predictive validity of polygenic scores for intelligence. The data for this meta-analysis were derived from nine independent samples, all composed of individuals of European ancestry from Western, Educated, Industrialized, Rich, and Democratic (WEIRD) countries.
In total, the sample included 452,864 participants, making it one of the largest investigations into the genetic predictors of intelligence to date. The included studies assessed intelligence using standardized psychometric tests, such as the Wechsler Intelligence Scale, which measure cognitive abilities in various domains, including verbal and non-verbal reasoning.
The researchers found that polygenic scores for intelligence showed a moderate but consistent ability to predict individuals’ performance on standardized intelligence tests. Across the nine independent samples analyzed, the polygenic scores were found to correlate at approximately 0.25 with IQ test results. This means that the scores explained about 6% of the variation in intelligence.
“This translates to a difference of ~4 IQ points, which, some research suggests, could be linked to differences in people’s education, employment, and income,” Oxley told PsyPost. “Genetic prediction could be really useful in research looking to disentangle how genetic and environmental factors work together in shaping us cognitively.”
But the researchers also uncovered significant variability in the predictive strength of polygenic scores across the included samples. Even after accounting for factors such as the type of intelligence being measured and the age range of participants, a substantial portion of this variability remained unexplained. This suggests that other unexamined factors, such as environmental influences or differences in the specific genetic variants included in the polygenic scores, may play a role.
“Critically, we also found that polygenic score predictions varied in strength across studies,” Oxley explained. “It was not clear from the data available why this is the case. Until we can explain why genetic predictions of intelligence are sometimes stronger and sometimes weaker, we won’t be able to use polygenic scores to help us understand people’s differences in intelligence at the individual level. More research will be needed to address this gap.”
The researchers also examined whether the type of intelligence being measured (e.g., verbal, non-verbal, or general intelligence) affected the predictive power of the polygenic scores. They found that verbal intelligence, which involves language-based reasoning and comprehension, was more strongly predicted by the scores compared to general or non-verbal intelligence. However, the polygenic scores were not able to differentiate strongly between other specific domains of intelligence.
“People’s scores on different types of intelligence tests are usually closely related,” Oxley said. “That is, people who score highly in maths tests also tend to do well on vocabulary tests. But we found that polygenic scores for intelligence didn’t predict all types of intelligence equally well. We ran a meta-regression, which revealed that polygenic scores predicted verbal intelligence significantly better than other types of intelligence – including general intelligence – with the weakest predictions seen for memory and crystallized intelligence (factual recall ability).”
“This could be partly because verbal intelligence is more likely to be inherited. People tend to have children with others who share traits in common with them, and verbal communication may matter more than memory or factual recall for most people. We didn’t see clear differences between predictions for other types of intelligence. This tells us that, while polygenic scores can predict people’s differences in general intelligence, they can’t yet be used to discern strengths and weaknesses in specific cognitive abilities.”
The study adds to a growing body of research showing that polygenic scores can provide valuable information about the genetic underpinnings of intelligence. But as with all research, there are limitations. First, it included only individuals of European ancestry from WEIRD countries, which limits the generalizability of the results. Research has shown that polygenic scores derived from European populations are less accurate when applied to individuals from non-European backgrounds.
The study also could not disentangle genetic influences from gene-environment interactions, where genetic predispositions and environmental factors influence each other. Gene-environment interactions occur when an individual’s genetic makeup influences their response to environmental factors. For instance, individuals with a genetic propensity for higher intelligence might seek out more intellectually stimulating environments, further enhancing their cognitive abilities.
“An important thing to remember is that genes do not determine intelligence: They correlate and interact with environmental experiences – like reading books and going to school,” Oxley told PsyPost. “Our differences in intelligence are shaped by genes and environmental factors, working hand-in-hand. However, it wasn’t possible in our meta-analysis to explore this interplay between genes and environments.”
“Another issue that we couldn’t explore in our study was age and generational trends. Research suggests that polygenic scores may predict intelligence more or less strongly at different ages or in different generations. The research data that was available to us here were not detailed enough to draw conclusions about age and generations, but we hope future studies will address these questions.”
Future research could address these limitations by incorporating within-family designs, which control for shared environmental factors, and by including more diverse groups in genome-wide association studies. Additionally, larger sample sizes and refined polygenic scores targeting specific cognitive domains might improve predictive accuracy and help disentangle the complex interplay between genetics and environment.
“We hope that our meta-analysis will inspire future genetically sensitive studies of intelligence,” Oxley said. “One question that we think is important to answer is why polygenic score predictions are stronger for some domains of intelligence and weaker for others. Another is elucidating how polygenic scores and environment come together to shape intelligence.”
“Behavior genetics is a rapidly progressing field. As we learn more, polygenic scores may become useful tools in applied contexts, for example in helping teachers and parents to identify children who might benefit from additional support. Before we can extend the use of polygenic scores to applied contexts though, it is vital to develop guidelines that ensure that genetic tools are not misused or misinterpreted.”
The study, “DNA and IQ: Big deal or much ado about nothing? – A meta-analysis,” was authored by Florence A.R. Oxley, Kirsty Wilding, and Sophie von Stumm.
