Genetics is a strange subject. Building upon the sequencing of the human genome in 2003, there is now a vast technical literature exploring the billions of steps in the double helix of DNA. It is not unusual for technical breakthroughs to generate controversial policy implications—we need think only of IVF or social media—but in the case of DNA its awesome potential is often accompanied by a nervous dismissal of any policy implications.
Finding 1271 education-associated genetic variants in a research study surely requires at least some speculation about the policy implications. In answer to a query about those implications, the study authors’ reply was “none whatsoever” (1). Cesarini and Visscher measure the heritability of educational attainment in an article whose technical merits would satisfy even the most fastidious research committee. They conclude by insisting their results are not relevant for evaluating changes to education systems (2). Cultural psychologist Steven Heine’s popular survey of the DNA literature asserts that “all human behavioural traits are heritable” but does not discuss policy implications so much as warn against them: it seems that we are anxious about DNA because we are “psychologically equipped to misunderstand it” (3).
This essay appears in October’s Quadrant.
Click here to subscribe
Such caution is understandable. In the 1960s and 1970s the consensus was for nurture rather than nature, and “it was dangerous professionally to study the genetic origins of differences in people’s behaviour and to write about it in scientific journals” (4). At the height of the US civil rights movement, amidst sometimes violent controversy, Arthur Jensen argued that genetics was responsible for the IQ gap between black and white Americans. In Britain, the Black Papers of the early 1970s were concerned with declining educational standards rather than specifically with genetics, but the tone was set by an article titled “Mental Differences between Children” by the geneticist Cyril Burt. Burt was subsequently accused of falsifying some of his statistical results, a scandal which did much to discredit hereditarian arguments in Britain. In recent years there has been some sympathy for the idea that he was guilty of carelessness rather than deliberate fraud. In the US, The Bell Curve by Murray and Herrnstein, a book of undoubted scientific integrity, postulated mostly hereditarian reasons for the differences in average IQ between black and white Americans, again drawing angry debate.
Into this minefield strides Robert Plomin with his recent book Blueprint. The subtitle, How DNA Makes Us Who We Are, wastes no time in softening up the reader for the provocative statements that come at a dizzying pace. For example, on average, differences in weight between people are largely due to inherited DNA differences. Parental divorce is the best predictor of children’s divorce but this association, easily interpreted as environmental, is actually due to genetics. About a third of the differences between children in their television viewing can be accounted for by genetic factors in their parents. Genetics is by far the major source of individual differences in school achievement, not the home environment or school quality (4).
This forthright approach has generated a response that goes back to the future with a vengeance. No less a journal than Nature doesn’t hesitate to unload, accusing Plomin of genetic determinism, playing fast and loose with the concept of heritability, and, however unintentionally, providing a road map for regressive social policy (5). Of course, this last accusation gives the game away, telling us more about the reviewer than about the book.
This article examines the claim that differences in genetics are the main determinant of children’s achievement in school. We focus on Plomin’s work, as he has most fully explored the educational implications of DNA, but we use other recent research where appropriate. Plomin notes that despite current policy interest in school quality, education is the field slowest to absorb the messages from genetic research. This is not so much through unawareness of its role but because of active hostility: “In the 1970s to mention genetics was just beyond the pale. Within the world of science and psychology, there is no longer any problem. But if you move out into other disciplines —[such as] education—genetics is still the devil.” (6)
Genetics in school achievement: the main findings
This reviewer is not a geneticist, and much of the professional jargon—alleles, polygenic scores, pleiotropy, genome-wide association, single-nucleotide polymorphisms (SNPs or “snips”)—means that the reader must look elsewhere for a comprehensive review of the science. We can report, however, that Plomin is scrupulous in adhering to the scientific principles of falsifiability and replicability; the book draws upon a mountain of evidence (he reminds us that the plural of anecdote is not data, an epigram that should be tattooed on the forehead of every so-called qualitative researcher); and the main statistical concepts (correlation and variance) are well within the grasp of anyone prepared to make a bit of effort.
People have no trouble with the idea of a genetic basis for characteristics such as height or eye colour. Many family-tree hobbyists are familiar with basic genetics through DNA tracing of their family origins. But whoever thought there might be genes for divorce or watching television? People know from their experience that school achievement tends to “run in families”, but this observation does not explain whether the cause lies in genetics or family nurturing. Most people would acknowledge that innate ability plays some part in schooling, but scepticism about an across-the-board genetic explanation is understandable. People are aware that bright children can come from parents of below-average ability and bright parents don’t always have clever kids.
To jump straight to the major finding from recent genetic work, school achievement is some 60 per cent heritable. Heritability is a measure of how well differences in people’s genes account for differences in their characteristics or behaviour. Put simply, more than half the differences between children in school achievement can be explained by inherited genetic differences.
The first question is what these numbers mean. A 60 per cent heritability for school achievement does not mean that 60 per cent of an individual’s school achievement was inherited from his or her parents’ genes. Nor does it mean that school achievement is 60 per cent heritable for each person. Heritability is fundamentally a statistical concept. If we have a sample of people whose school achievement is compared with their genes, we will observe individual variations in that relationship. Those variations can be summarised in a single quantitative statistic measuring the extent to which differences in achievement can be attributed to genetic variations. Heritability of 60 per cent means that, on average in that sample, three-fifths of measured differences in school achievement can be attributed to differences between people in their inherited genes. The remaining 40 per cent is due to other, perhaps unknown, reasons such as life experiences or “nurture”. (Note that heritability is not 100 per cent, which would imply that all the variability in achievement comes from genetic differences. This arithmetic explains why some individuals in the sample may be very successful at school despite coming from not-so-bright parents. The converse is also true.)
The second question is how these estimates are derived. Much of the early work on heritability was carried out through studies of twins and adoption. It has long been known that identical (monozygotic) twins come from the same fertilised egg and therefore are 100 per cent similar genetically. Twins separated by adoption early in life share nature but not nurture. With the effect of genetics controlled through the accident of conception, twins have long provided an experimental design for measuring the relative importance of nurture and nature.
Identical twins reared apart are rare, but the same principles have proved to be flexible tools with generalisable findings. In Sweden a large sample comparing seven types of sibling pairs who differed in their genetic and environmental relatedness showed a pattern of increased correlation between genetic relatedness and years of schooling. For adoptees reared together but sharing no biological parents (thereby measuring only the effects of nurture) there was only a low correlation with school achievement (7).
Twin/adoption studies are still part of the genetic toolkit, but the sequencing of the human genome in 2003 revolutionised the experimental approach. Now there was the possibility of using DNA techniques to locate which genes were responsible for observed differences in schooling, health or behaviour.
This has shown that there is no such thing as a unique “candidate gene” associated with this or that behavioural trait. What we find instead are thousands of small genetic differences, each of barely measurable effect. Detecting the influence of such tiny genetic effects across the genome (genome-wide association, or GWA) requires very large samples. In less than ten years, sample size has rocketed from barely 8000, with no genetic “hits” for education, to over one million in a recent study, resulting in 1271 significant genetic associations for education (1).
It also became possible to combine these minuscule effects into composite polygenic scores. Polygenic scores are the holy grail of genetic analysis. This is because they attempt to identify the specific genetic markers responsible for some trait. Each individual has a different polygenic score, so a major goal is to estimate genetic propensities specific to the individual and thereby predict outcomes. Lee et al do not estimate individual scores in their sample but divide their scores into five groups, or quintiles (1). Just as with the Swedish study cited earlier, they find a steady rise in school achievement (in this case the prevalence of college completion) from the lowest quintile (12 per cent prevalence) to nearly 60 per cent in the quintile with the highest polygenic score. In Plomin’s own research he found that the average GCSE score (the examination taken at the age of sixteen in Britain) increased steadily as the education attainment polygenic score increased.
What are the policy implications for education?
The balance of evidence suggests that Plomin and other researchers have made their case that a large part of variation in school achievement is explained by inherited DNA. The insight that twins can be used as a controlled experiment for the role of nature/nurture looks slightly quaint when compared with the awesome technology of genome sequencing, but the twin studies (Burt’s carelessness or fraud aside) have an impressive record of consistent findings. They are supported, not contradicted, by the findings from polygenic scores.
The consequence, Plomin concludes, is that DNA is “the best predictor we have of a child’s years of education, even better than the environmental effect of family socioeconomic status”. Indeed, the arithmetic of heritability understates the role of genetics. Heritability of around 60 per cent implies that “nurture” or environment accounts for the remaining 40 per cent. But what we take to be environmental influences can often have a strong genetic component—”the nature of nurture”. For example, parents’ reading to children may affect how well the children do at school. This effect, apparently due to nurture, may in fact be at least partly a consequence of genetics. Parents and their children are 50 per cent related genetically, so parents who like to read tend to have children who also like to read or be read to.
Genetic screening and predicting school achievement
The most obvious implication from genetic research is that it may become possible to predict an individual’s school achievement from his/her polygenic score. We are still a long way from doing that with any degree of confidence. Using polygenic scores to compare group averages is one thing: predicting the achievement of an individual is quite another. There is a range (standard error) around the average of each group and substantial overlap in achievement between genetic groups—as we would expect when heritability for school achievement is 60 per cent, not 100 per cent.
Plomin acknowledges that individual prediction can only be probabilistic, not a certainty, but it is fair to describe him as an enthusiast for early genetic testing of individual children. This would help spot learning deficiencies and allow for earlier intervention. He rightly notes that interventions work better the earlier you start, so children should be genetically screened as early as four years old to predict future problems in their education. Tests such as NAPLAN are intended to be diagnostic and forward-looking, but in practice are based on recent performance. Why not use the technology of DNA to screen children before they even start school?
This line of argument leads to the notion of “genetically sensitive” schools. In such schools a child’s genetic information would be used to tailor the curriculum and teaching to create a system of “personalised learning”. Outlandish as this seems, all school systems already have some “genetically sensitive” features. Classes are often streamed; children with poor results are likely to repeat a grade; and those with learning disabilities may be taught by specially trained teachers using different pedagogies.
None of this alters the outcry that would result from any attempt to introduce genetic testing as a deliberate act of school policy. It is all too easy to imagine the dystopian argument that genetic testing would be used to deny educational opportunities to various ethnic or socio-economic groups. The 1997 film Gattaca about a world divided into “valids” with ideal genomes and a genetic underclass of “in-valids” would be quoted endlessly. But we don’t need pop culture or a conspiracy theory about eugenics. The more practical concern is that such a policy could worsen what has already been a serious issue in Australian education—a complacent acceptance of unacceptably low standards. Genetics does not predict what the limits might be on school achievement, but schools with low achievement from an enrolment with predominantly low polygenic scores could claim they are doing as well as can be expected from the raw material they have available.
Plomin is probably correct in forecasting that “genetics screening tests are going to happen … It is not something we can do [right now] but it will happen.” Eventually it will become possible to predict which children are going to find it more difficult to learn and we will be able to identify the genes involved.
Rethinking equality of opportunity
Genetics has always seemed antithetical to equality of opportunity. People who get lucky in the genetic lottery have a better chance of doing well at school and this seems to violate the idea that everyone should have fair and equal access to education. Those lucky children can in due course enjoy high occupations and earnings and pass on educational advantages to their own children, entrenching inequality of opportunity. Given that there are elements of the “straw man” in such a description, the argument is common enough for it to come as a shock that DNA is not the enemy of equal opportunity.
The starting point for this paradox is the commonplace observation that equality of opportunity does not translate into equality of outcomes. If educational opportunities were the same for all children, school achievement would still vary because genetic differences would remain. If outcomes are a function of genetics and environment, simple arithmetic tells us that reducing environmental obstacles leaves genetics as the main determinant of outcomes. In short, the more we improve opportunity, the greater the role of genetics.
Paradoxical though it seems, the higher the heritability of education in a society, the more equal is the level of opportunity. Plomin concludes that the 60 per cent heritability of school achievement in Western countries suggests there is substantial equality of opportunity. Genetics is not antithetical to equal opportunity: it shows how well we are doing.
Probably we are doing even better than a heritability of 60 per cent would imply. Through the effect of “nature of nurture”, at least some of the remaining 40 per cent is strongly influenced by genetics. Parents enjoying high occupations because of educational advantages may indeed pass on those advantages to their own children. But the relationships here are substantially heritable, not environmental. As soon as we start talking about parents and their children, we are talking of a relationship largely driven by genetics. Parent-child correlations in education achievement are mediated by heritability, and high heritability is actually an indicator of equal opportunity.
Does education make a difference or is it all in the genes?
Writing against a British institutional background that translates comfortably to Australian education, Plomin found that differences between schools did not make much difference to student achievement. Quality ratings by the influential UK Office for Standards in Education (Ofsted) explained less than 2 per cent of variance in secondary achievement (GCSE scores) after correcting for students’ prior achievement in primary school. Similarly, students in selective schools performed negligibly better than those in non-selective schools, again after controlling for prior achievement.
Plomin reached this conclusion by allowing for what’s known as selection bias. When we compare achievement in different schools, it’s important to test whether school A performs better than school B because it provides superior education, or whether that result is confounded by other factors. The most obvious example is that a school’s higher achievement may be due mainly to having brighter students. Data permitting, it is often possible to control for student ability by including a variable for prior academic attainment. This corrective procedure compares schools only after “netting out” ability differences between students.
Differences between students in public and private schools, or selective and non-selective schools, are often thought to be environmental in origin, but prior ability and intelligence are substantially heritable. The selective process arises for genetic reasons. To put it crudely, it’s apparently all in the genes. Selective schools choose higher-ability students. Even for comprehensive state schools, family socio-economic status may play a role in the type of school a parent selects. Schools located in more affluent areas attract better students. Ofsted’s quality ratings of schools (and NAPLAN in Australia) contribute to, and reinforce, this genetic sorting because parents use such ratings to get their children into a “good” school through “postcode shopping” and the like.
The outcome is a process in which schools select for higher ability or where parents do the selection, with students of greater ability clustering in a hierarchy of schools. In both cases such schools demonstrate high academic achievement, but Plomin argues that this is due to the genetically driven abilities of the students. The school itself contributes nothing special in the way of better academic outcomes. Save your money is Plomin’s advice to parents, because “expensive schooling cannot survive a cost-benefit analysis on the basis of school achievement itself”.
It’s doubtful if this is good advice. There is no reason to doubt the claim that differences in genetics are the main determinant of children’s achievement in school, and this factor dominates the selection process, but there is more to a good school than just good genes. It’s implausible, to say the least, to suppose that schools do not add educational value beyond reflecting the contribution of DNA differences between students. Any competent teacher understands that you must lift your game if you have a highly capable class.
The problem is that correcting for selection bias can in practice have perverse results. Often it succeeds only in over-correcting the statistical measures. Whether we are talking selection by schools or by parents, exam results are the visible and measurable outcome of a cluster of complementary attributes of a good school. Among these attributes are not just student ability but school disciplinary policy, amount of homework, academic streaming, socio-economic standing of the parents, high expectations of performance, the beneficial role of student peer groups, more effective teachers and leadership by the principal.
All these attributes are part of the complex relationship between ability and school achievement. Hold student ability constant and you implicitly hold these complementary attributes constant, thereby seriously understating genuine sources of advantage by schools. These attributes are the very mechanism that distinguish an educationally effective school from a bad school and should be allowed to vary in the statistical work. Control them, and we rule out the very thing we want to test. Correcting for student ability, without otherwise measuring these factors, means that we are attributing to student ability effects which really are part of the school environment.
It’s not surprising that genetics has a mixed record of success when dealing with education policy. The underlying science of DNA has made astonishing progress in the last decade or two, and Plomin is surely right in suggesting that education has scarcely begun to absorb the messages from genetic research.
Some of those messages are challenging, as in early-age genetic screening and the prediction of school achievement. The most appropriate public policy in the short run would be to leave it to parental choice. Testing one’s ancestry through DNA is already a thriving business. It is increasingly possible to have one’s entire genome sequenced at moderate cost. If we allow both the market and the technology of genetic screening to develop, we will learn to compile effective interventions that make genetic screening not a deterministic dead-end but a useful supplement to existing systems such as NAPLAN.
Equally challenging is the idea that the classic inter-generational inequality of opportunity is to a considerable extent explained by genetics rather than environmental factors. Plomin argues that in a society with universal basic education, teaching literacy and numeracy in the early school years largely erases environmental disparities. This leaves genetics as the primary cause of differences between children in those skills. Indeed, the higher the heritability of education in a society, the more equal is the level of opportunity.
The main policy issue is that we need to ask where we should look for further improvement in opportunity. If schools differ in their success in teaching literacy and numeracy in the early school years, then environmental differences will persist. Much has been written in recent years about the poor performance of many Australian schools. Much of this writing has come from the perspective of improving school quality for reasons of human capital and economic productivity. To this we must now add the objective of improved opportunity.
One of the clichés of the school performance literature is that improved performance is largely a matter of raising standards of the lowest-performing schools and students so that they reach their potential. In Australia indicative evidence suggests that over 14 per cent of primary schools obtain average NAPLAN scores that put them at or below the national minimum standard. Policies aimed directly at schools with large numbers of students not achieving national minimum standards are vital.
It is also worth noting that further improvement in equality of opportunity from environmental sources will bring us face-to-face with the genetic differences that remain. Inequality from environmental reasons will be replaced by inequality of outcomes arising from genetic differences between people.
The argument that the school itself doesn’t make a difference to achievement over and above the role of school ability exemplifies the gulf between genetics and education policy—and the gulf is not just one way. Genetics has always had a bad rap in education. It is true too that genetics has made outstanding progress in recent years. That progress is nicely seen in the work claiming that there is little difference in school achievement after controlling for selection factors. The genetic part of that research, with its use of polygenic scores, is work of considerable sophistication. It is a pity, therefore, that the research does not reflect progress over recent decades in what we have learned about the complex and varied determinants of educational performance.
Finally, it is worth noting that the brute force of sample size and today’s computing power can produce genetic correlations that pass muster statistically, but the causal pathways between DNA, the brain, and behavioural traits remain only poorly understood. Plomin claims this as an advantage: we can use inherited DNA differences to predict individual difference without knowing anything about the myriad pathways connecting genes and behaviour. This is an unlikely advantage. We can use huge samples to predict on purely statistical criteria that differences in school achievement derive substantially from inherited DNA, but size isn’t everything. Without an explanatory model, DNA results will remain something of a statistical “black box”.
Ken Gannicott was previously Professor of Education at Wollongong University. He now works as a consultant and has undertaken many international assignments.
- James J. Lee et al. (2018), “Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals”, Nature Genetics.
- David Cesarini and Peter M. Visscher (2017), “Genetics and educational attainment, review article”, Science of Learning 2:4.
- Steven J. Heine (2017), DNA Is Not Destiny, W.W. Norton and Company, New York.
- Robert Plomin (2018), Blueprint, How DNA makes us who we are, Penguin Random House, UK.
- Nathaniel Comfort (2018), “Genetic determinism rides again”, Nature, September 561.
- Andrew Anthony (2018), “So is it nature not nurture after all?”, The Guardian, 30 September.
- Rietveld, C. A. et al. (2013), “GWAS of 126,559 individuals identiﬁes genetic variants associated with educational attainment”, Science 340, 1467–71.