In this post, CitiGen researcher Sarah Abel tries to get to the bottom of some of the claims surrounding a recent study about genetics and educational attainment.

Last month, a number of articles and opinion pieces in the British press reported on a new study entitled “Differences in exam performance between pupils attending selective and non-selective schools mirror the genetic differences between them“. Below is a selection of the headlines:

The Telegraph, 23 March 2018
The Guardian, 23 March 2018
Financial Times, 24 March 2018
The Times, 26 March 2018

The idea that human intelligence – like almost all human behaviours – is to some extent influenced by genetic factors is no longer controversial among experts in the field of intelligence. Nonetheless, research on genetics and educational attainment (often used, like IQ, as a measure of intelligence) has frequently aroused controversy, both in and outside of academic spheres. Media discussions of studies such as this one often follow a similar pattern, with analyses moving quickly from overviews of the research findings to heated speculation about how they relate to current and future educational policy. Within these analyses, reports typically zoom in on any signs that scientists are claiming to have identified a genetic component to intelligence, and interpretations of this evidence frequently become polarised, with the importance of genetic factors sometimes dismissed outright, and other time grossly exaggerated by commentators.

The potential for studies involving human subjects to inform important public policy decisions, combined with a long history of misuses of human research to support racist, classist, and eugenic ideologies and programmes, suggest that it is neither unwise nor unwarranted to subject such research to rigorous scientific critique, as well as to lively debates about related ethical and political implications. At the same time, this exercise becomes difficult and self-defeating when commentators begin with ready-formed and often intransigent notions about the extent to which genetics can or cannot explain aspects of human behaviour.

In light of these considerations, this post aims to clarify what the original study actually said (and didn’t say) about the influence of genetic and other factors upon educational attainment among schoolchildren, and to explain how this fits in with the current academic consensus on the topic.

What the researchers did

The study was conducted by a group of researchers from King’s College London, led by psychiatrist Emily Smith-Woolley and behavioural psychologist and geneticist Robert Plomin. Plomin himself is internationally recognised for his pioneering research into the impacts of genetic variables upon complex behavioural traits. In their introduction to the paper, published in the journal Science of Learning, the group stated their aim to assess the respective roles of DNA and selective schooling – among other factors – upon educational attainment in British students at GCSE level (the standardised exams taken at age 16 in the UK).

The researchers gathered socioeconomic, genotype, and educational attainment data (GCSE scores in maths, English and science) from 4818 schoolchildren. Participants were categorised according to the type of school they attended: 1) non-selective schools (i.e. state-funded schools with no entry requirements); and 2) selective schools – including grammar schools (state-funded, with entrance based on exam scores), and private schools (fee-paying). They were also attributed a genome-wide polygenic score (GPS), developed in a previous study, which takes into account thousands of genomic markers, found to be associated with individuals’ years spent in education (referred to as EduYears – another metric for assessing intelligence).

The authors’ approach was to first compare the average polygenic scores of students at selective vs. non-selective schools, and try to explain any variation in EduYears GPS between the cohorts. Then, they aimed to explain any disparities in average exam results among selective and non-selective schools, assessing to what extent these could be accounted for by the quality of schooling vs. other factors, including genetic variability.

Findings

The researchers reported the following initial findings regarding students’ polygenic scores:

  1. On average, students at non-selective schools scored significantly lower on the EduYears GPS than students at selective schools. As Plomin stated an interview with Adam Rutherford for the BBC podcast Inside Science: “Students from selective schools are three times more likely to be in the top group of this polygenic score than those from non-selective schools“.
  2. The EduYears GPS correlated positively with several of the factors that determine students’ entry into selective schools, namely: family socio-economic status (SES), prior ability, and prior achievement. The researchers linked this observation to the previous one as follows: “students with higher polygenic score for years of education have, on average, higher cognitive ability, better grades and come from families with higher SES, and these students are subsequently more likely to be accepted into selective schools“.

Regarding the average GCSE scores of students at selective schools vs. non-selective schools, the study data showed a roughly 7.1% disparity, with students at grammar and private schools gaining higher marks on average than students at state schools. Plomin, in conversation with Rutherford, put it like this:

The difference between selective and non-selective schools is a whole grade, it’s like an A versus a B, on average. This is well known. But people assume that’s environmental – it’s easy to do, I mean if you’re forking out all that money to send your kid to a private school, you want to bet it’s because your kids going to get a better education.

To test this idea (i.e. that increased educational attainment among students in selective schools is primarily due to the schools’ quality of resources and teaching) the researchers controlled again for the other factors that influence the selection of students for grammar and private schools (family socio-economic status, prior ability, and prior achievement). They also controlled for the influence of genetic variants associated with years in education, which were found in a previous study to have a small impact on GCSE score variance. These steps yielded the following results:

  1. Once factors involved in student selection and the EduYears GPS were controlled for, the variance between exam scores in selective and non-selective schools dropped from 7.1% to 0.5%. In the authors’ words: “We find that almost all of the selective school advantage in GCSE can be explained by family SES, achievement, ability and EduYears GPS. After controlling for these factors, going to a grammar vs. a state non-selective school is associated with a mean GCSE grade increase of just 0.026 of a standard deviation and for private schools, 0.070 of a standard deviation”.

Reactions and interpretations

As the headlines displayed at the top of this post indicate, the different newspaper reports of the study’s outcomes all captured what the researchers themselves saw as their main takeaway conclusion: that sending children to grammar and private schools does not appear to have a significant impact in terms of improving their academic potential.

Yet unlike the original article, many reporters and commentators went on to affirm that genetics was the principal factor behind differences in academic achievement among students. For instance, Matt Ridley wrote in the Times:

A new study finds that selective schools add almost nothing to the exam results of students, because the advantages teenagers come out with are mainly ones they arrived with, and are for the most part genetic.

Likewise, Camilla Turner wrote in the Telegraph:

[Researchers] found that children who attend grammar or private schools are more likely to do well in exams – but this is largely down to their genes, rather than their school environment.

Going back to the scientific article, the researchers’ original statement was that: “almost all of the selective school advantage in GCSE can be explained by family SES, achievement, ability and EduYears GPS”. Among these factors, the study authors noted that the EduYears GPS in fact had the smallest impact on average GCSE grades between school type. In fact, contrary to what was reported in much of the national press, we could say that the variance currently explained by the EduYears GPS is so low that it is rather difficult to quantify (although Plomin has predicted that this value will rise as more genetic variants with associations to measures of intelligence are discovered and added to the algorithm). Instead, prior exam achievement had the biggest impact, followed by family SES and prior measures of cognitive ability.

Yet, this is not to say that the EduYears GPS is the only one of these variables that could be influenced by genetic factors. As Plomin stated to Rutherford:

The gist of the paper is to say […] well, selective schools – what are they selecting children on the basis of? It’s ability, achievement, and socioeconomic status. And as a geneticist, I say: those are highly heritable characteristics – each one of them. So that, maybe, the mean difference between schools and GCSE scores, and the DNA differences, could be due to these selection factors – and that’s exactly what we found.

Heritability has been defined in the context of behavioural psychology as “the proportion of variability in a phenotype that is ‘accounted for’ […] by variation in genotype”. Yet here, Plomin’s reference to socioeconomic status as a “highly heritable characteristic” suggests he is using the term in a slightly looser way, corresponding to things that can be inherited from generation to generation – not just traits that are heritable genetically. This distinction is well illustrated in this example by bioethicist Iain Brassington:

[W]ealth, like genes, is passed down the generations. And wealth tends also to correlate with things like education, exposure to certain cultural phenomena, and the like.

Indeed, studies indicate that the other two selection factors mentioned by the researchers (prior educational achievement and prior cognitive ability) also have a tendency to “run in” families. Of course, neither of these are inherent traits. They are metrics that begin to be calculated at an age when most children have already had several years’ worth of exposure to potentially divergent educational experiences within their family life. As Brassington continues:

It would not be a huge surprise to find that someone born into a well-read household from an academic background, whose family has plenty of money to spend on things like trips to the theatre or museums in other cities, grew up to be highly articulate and able to engage fluently with all manner of intellectual pursuits without batting an eyelid. By contrast, if you’re born into a household in a deprived area, in which you have to share a bedroom with a sibling or maybe more, and in which money is tight, the chance that you’ll be able to go to somewhere quiet and read books is reduced. You may not have anywhere quiet to go, and you may not have the books. You may not be taken on trips to museums because of the transport costs.

Brassington’s point is not to dismiss the idea that genetics can be a factor in influencing intelligence. Rather, it is to point out how difficult it is, in practice, to extricate the influence of genetic heritability from the impact of other circumstances, opportunities and experiences that are effectively transferred from generation to generation by virtue of social and legal structures that shape our family lives and societies in general.

Indeed, rather than hypothetically separating “genetic” from “environmental” factors, Brassington’s examples support the idea that a better way of imagining these phenomena would be to view diverse biological and social factors as tightly intertwined and, in many cases, mutually reinforcing from generation to generation. For instance, in another example, Brassington points out that certain studies have shown cognitive ability to be related to exposure to parasites, explaining:

[I]f your body is having to put energy into fighting off infections, it’ll have less energy to devote to things like cognitive development; and so if you’re exposed to a lot of pathogens, especially in early life, you might end up less smart than you otherwise would be.

Brassington also notes that “exposure to infectious illness, and lacking the means to do much about it, correlates with poverty”. In this instance, low SES and environmental biological factors may combine to have an impact on a child’s cognitive development, leading to poor school performance, which could in all likelihood have a long term impact on her job prospects, future earnings and social mobility. If nothing changes from one generation to another, scientists studying a case like this one might infer that this family suffers from “low intelligence”, and conclude that this is the cause of their long term impoverishment. On the other hand, examining the interlinking effects of these biological, environmental and social factors would lead the researchers to a different conclusion: that poverty is causing members of this family to consistently score low on measures of cognitive ability, leading them to appear less intelligent than they could do in altered circumstances. Or, as Brassington puts it: “Bluntly, someone might look stupid, but simply be stupefied”.

Human beings or human becomings?

It is also important to get away from the perception that our genetic inheritance has fixed, inevitable impacts in terms of shaping our bodies and behaviours. Since the sequencing of the human genome in the early 2000s, scientists have come to understand that only in very rare cases can particular phenotypes be linked directly and simply to individual alleles.

Rather, genome-wide association (GWA) studies are now seen as the best way to begin to discover how patterns of genetic variation can be linked to particular phenotypes. This type of research involves surveying tens or hundreds of thousands of single nucleotide polymorphisms (SNPs – “letters” in genetic code that vary among individuals and groups) using DNA collected from thousands of research participants, and identifying SNPs that correlate with particular phenotypes. While each SNP taken individually may have an almost insignificant association to a given phenotype, when studied collectively, researchers can start to see more significant patterns of correlation.

At the same time, experts emphasise that we are still unlikely to find exclusively genetic causes for many phenotypes. As Eric Turkheimer, Kathryn Paige Harden and Richard E. Nisbett of the Genetics & Human Agency project wrote in a recent article for Vox:

Modern DNA science has found hundreds of genetic variants that each have a very, very tiny association with intelligence, but even if you add them all together they predict only a small fraction of someone’s IQ score. The ability to add together genetic variants to predict an IQ score is a useful tool in the social sciences, but it has not produced a purely biological understanding of why some people have more cognitive ability than others.

It is also worth remembering that statistical correlations between genotypes and phenotypes do not necessarily indicate a direct or causal relationship – a rule of thumb that Graham Coop has helpfully illustrated with a hypothetical example about polygenic scores and tea-drinking. Or, as Ewan Birney, director of the European Bioinformatics Institute, pointed to out Adam Rutherford regarding the Smith-Woolley et al. paper:

When we’re talking about genetic difference and educational attainment […] there could be a variety of different hypotheses or underlying bits of biology. […] There’s probably a set of effects which are related to taking exams or related to taking tests, but aren’t really fundamental to what we think of as intelligence. And here’s a made-up example: maybe people with smaller bladders, who have to go to the loo more, they will end up not performing so well in exams. Now – I have no idea whether that’s true or not, but if it was true and there was a genetic effect there, we would see it in these kinds of [polygenic] scores.

Meanwhile, in their piece, Turkheimer, Harden and Nisbett go on to signal another crucial point. Although many non-geneticists still often tend to think of innate, immutable characteristics when they hear the term “heritable factors”, geneticists now agree that most heritable traits – though partly genetic – are also highly modifiable:

[H]eritability, whether low or high, implies nothing about modifiability. The classic example is height, which is strongly heritable (80 to 90 percent), yet the average height of 11-year-old boys in Japan has increased by more than 5 inches in the past 50 years. Average IQ scores are increasing across birth cohorts, such that Americans experienced an 18-point gain in average IQ from 1948 to 2002. And the most decisive and permanent environmental intervention that an individual can experience, adoption from a poor family into a better-off one, is associated with IQ gains of 12 to 18 points.

Once again, this observation points away from the classic notion of a divide between “genes” and “environment”; “nature” and “nurture”; the “biological” and the “cultural”. It also contradicts the idea – put across by at least one reporter of the Smith-Woolley et al. study – that it is “our genes” that “shape the environments we make for ourselves and our children”, rather than a combination of political choices and historical socio-economic structures, over which we as citizens have some degree of influence and capacity for change.

Anthropologists Tim Ingold and Gísli Pálsson have considered this problem at length, and in their recent work Biosocial Becomings they offer an alternative to the neo-Darwinistic viewpoint that humans are inescapably shaped by the cumulative evolutionary effects of “genes” and “culture”. In his introduction to the book, Ingold proposes that breaking away from this approach means dismantling some of the fundamental views we hold about ourselves as a species:

And the first step in doing so is to think of humans, and indeed creatures of all kinds, in terms not of what they are, but of what they do.

Another way of putting this […] is to think of ourselves not as beings but as becomings – that is, not as discrete and pre-formed entities but as trajectories of movement and growth. Humanity, we argue, does not come with the territory, from the mere fact of species membership or from having been born into a particular culture or society. It is rather something we have continually to work at, and for which, therefore, we bear the responsibility.

Changing the paradigm from human beings to human becomings is an act of liberation. It can underline our capacity to influence the future of our societies, including the opportunities of both our children and others’ – rather than viewing our collective lives as beholden to our so-called “selfish genes”.

Yet, Pálsson and Ingold point out that this shift should not lead us to reinforce once more the divide between the biological and the social (what we are and what we do); instead, they suggest taking away the divide altogether, by seeing life as biosocial. In their words:

The life of becoming […] could be compared to a hempen rope, twisted from multiple strands, themselves twisted from multiple fibres, each in turn twisted from its cellular and molecular constituents. It could, in principle, be examined close up or from afar, microscopically or macroscopically. But at every level of resolution we find the same complexity, the same intertwining of threads, the same metabolic exchange. Like the rope, the becoming is biological all the way up, and social all the way down.


If you enjoyed this post, sign up below to have future blog posts from CitiGen delivered straight to your inbox!

[Thumbnail image credit: Alex Nabaum]