Reading ability: nature or nurture?

For whosoever hath, to him shall be given, and he shall have more abundance: but whosoever hath not, from him shall be taken away even that he hath.

Matthew, 13:12

The Matthew Effect has become something of a truism. Those with find it easy to acquire more, whereas those without are trapped into a vicious cycle of poverty and disadvantage. Clearly this is a matter of social injustice: if only we could ensure that all were treated equally then we could do away with such asymmetry. This is something I’ve been particularly interested in ever since hearing Geoff Barton refer to Daniel Rigney’s book, The Matthew Effect: How Advantage Begets Further Advantage. In it Rigney summarises the work of Keith Stanovich and discusses how reading ability depends on social advantage.

This seemed to make perfect sense and I accepted it uncritically. (My post on this is the second result suggested by Google when you search for the Matthew Effect.) But as ever, comforting, convenient answers aren’t always as correct as we’d wish them to be.


Following a recommendation from Andrew Sabisky I’ve been reading Kathryn Asbury and Robert Plomin’s G is for Genes: The Impact of Genetics on Education and Achievement, and it has presented some new and troubling (to me) information. Asbury  estimates that heritability accounts for between 60-80% of reading ability. (p. 24) This leaves as little as 20% down to the nurturing effects of our environments. While it’s not really possible to point out such percentages in individuals, twin studies across three continents have allowed researchers  to estimate the genetic and environmental influences on reading for population groups.

Admittedly, this is horribly oversimplified, and Asbury says:

Genetic variation exists from the moment we are born, but is multiplied and magnified as our genes interact with each other and our environment. It is likely that some environmental effects are hidden within our heritably estimates because they are effective indirectly, via their interplay with genes. (P. 27)

The example offered to illustrate this interplay comes from this study. Researchers compared children’s reading ability in Colorado, New South Wales and Scandinavia. In each of these territories there are different environmental factors at play. In both America and Australia children are required to attend school from the age of 5. In New South Wales, children attend school from 9-3 every day and the state mandates that 35% of this time be devoted to centrally directed literacy instruction, but in Colorado, children only attend kindergarten for 3-4 hours a day and the curriculum is left entirely in the hands of individual schools. In Scandinavian countries children do not begin reading instruction until the age of 7. Unsurprisingly, researchers found enormous differences between the interplay of genetic and environmental factors:

Screen Shot 2014-07-14 at 10.05.56

But, after each of the different population groups had received 1 year of reading instruction, differences largely disappeared:

Screen Shot 2014-07-14 at 10.06.05* Please note: these figures have been rounded up

When environmental factors are broadly similar, genes account for the vast majority of the difference in reading ability. And very interestingly, shared environment (growing up in the same household with the same parental influences) practically vanishes as a source of influence.

This is deeply counterintuitive. As Asbury puts it, “More school – that is, more environmental input – leads to greater genetic influence rather than greater environmental influence.” (p. 28) As children’s environmental influences become more similar, genetic differences become more noticeable.

Reading is unnatural and, genetically speaking, a very recent development in human communication. We all pick up speech without formal instruction, but nobody just learns to read, no matter how genetically predisposed they might be having a high reading ability. So if we look at samples of children before they have begun to receive formal reading instruction at school it seems reasonable to infer that the greater part of their reading ability comes from their environments. Parents who value reading and read to their children are more likely to have children who can read. The ad hoc reading instruction in Colorado results in some non-shared environmental reading influence whereas the lack of any formal instruction in Scandinavia before the age of 7 suggests that reading ability is most likely attributable to shared environmental influences.

But, education is the great leveler. When all children have received similar reading instruction, the differences between them are explained by genetic influences, resulting in a bell curve with a normal distribution of reading ability:

bellcurveThis goes to show the huge importance of non-shared environmental influence (schooling) has. While the differences between children are mainly accounted for by genetic influences, education accounts for the fact that children learning to read is not left to chance. Although schools cannot eliminate these genetic differences, they can attempt to move the entire bell curve further to the right. “Genes, and therefore human potential, cannot grow in a vacuum.” (P. 30)

This is, I think, cause for some optimism. All children are different. We each have our own gifts, talents and potential, and no two people are exactly alike. But school provides – or should provide – the opportunity for all children to excel to the best of their ability. We cannot all be geniuses, we cannot all be wonderful readers, but we can all be better (and often significantly better) than we currently are. This then is the power and the possibility of the growth mindset. Blaming or rewarding children for their genetic inheritance is clearly ludicrous, but demonstrating that hard work, effort, practice and a willingness to learn from making mistakes allow us to achieve our genetic potential is a much fairer approach to take.

I’m only half way through G is for Gene and am finding it a challenging read; there are some aspects and conclusions of which I’m sceptical, but books which challenge our assumption are the only ones really worth reading.

Related posts

Grit and growth: who’s to blame for low achievement?
Why we disagree: the purposes of education
The Matthew Effect – why literacy is so important

Also, Assessment, Standards and the Bell Curve from Tom Sherrington

49 Responses to Reading ability: nature or nurture?

  1. Jack Marwood says:

    Excellent food for thought, as always David. Reading the opening of Katherine Asbury and Robert Plomin’s G is for Genes on Amazon ( I’m pleased to see them referring to school as the Icing on the Cake… I’m intrigued, however about the terms used.

    “When environmental factors are broadly similar, genes account for the vast majority of the difference in reading ability. And very interestingly, shared environment (growing up in the same household with the same parental influences) practically vanishes as a source of influence.” How were genes and ‘shared home environment’ isolated as factors?

    A further thought is the idea that reading is somehow influenced by genetics, given that widespread reading in an advanced country like England is only a few hundred years old. Many people whose relatively recent family would not have been able to read seem to be advanced readers. Do the authors discuss this issue?

    • David Didau says:

      1) They do go into how variables were isolated but I’m not sure I completely follow the mathematics. For the moment I’m taking it on trust 🙂

      2) There is no such thing as a gene for reading. But, there are a range of genes that have a bearing on reading ability: vision, hearing, imagination, memory etc. All of these attributes accrue to make us relatively more or less able able at reading.

      • Jack Marwood says:

        Thanks David.

        1) Fair enough! I’d be interested to know, since each parent clearly supplies both some genes and some home environment; either way, it seems that the broad thesis is that school doesn’t add a huge amount on top *when compared to the influence of parents* – that huge 80% you note above.
        2) So those who have genetic disposition to be reasonably successful human beings (and largely those who have developed and benefited from reading, I’m guessing), are reasonably good at reading? That makes sense… It does all get close to some kind of genetic determination, doesn’t it?

        • David Didau says:

          “It does all get close to some kind of genetic determination, doesn’t it?” It does seem as though that’s the case. Interestingly though, the influence of parents is largely restricted to their geneses as adoption studies and twin studies have shown. I’m not sure yet what to do with this information although on the positive side it does rather undermine the Larkin hypothesis: “They fuck you up, your mum and dad…”

          • MaggieD says:

            “So those who have genetic disposition to be reasonably successful human beings (and largely those who have developed and benefited from reading, I’m guessing), are reasonably good at reading? That makes sense… It does all get close to some kind of genetic determination, doesn’t it?”

            I think you might be coming close here to saying that you need to be intelligent to be a successful reader. Which isn’t the case. You just need the requisite processing mechanisms (sorry, I’m not sure if that is the right way to describe it but I hope youundestand what I mean).

            Setting aside the fact that ‘successful’ can have multiple interpretations I think that are probably a considerable number of people around who are illiterate or semi literate yet who appear to be leading successful lives.

            It may appear from interpreting Plomin’s research that succcess at reading might be something of a genetically based lottery, but the neccessary processing mechanisms (?) seem to be common to all but a very small percentage of people; most people of whatever ethnic origin seem able to learn to read, though some may need more intensive instruction than others.

          • David Didau says:

            No, I’m saying the exact opposite. In order to read you need to be taught – intelligence has nothing to do with it. But it would appear that reading ability, once we have been taught how to decode, distributes across a normal bell curve just like every other human ability or characteristic, and the main factor in determining this distribution is genetic inheritance.

            My reading of Asbury & Plomin is also far removed from yours – they very clearly say that access to instruction is vital to being able to read, so that that extent at least it is a lottery – if you live in a part of world with little or no access to education you are unlikely to learn to read regardless of your genes. They agree that close to 100% of people can learn to read and that this should be a very reasonable goal of education. Does that make sense?

    • Andrew Sabisky says:

      As David says, there are clearly no “genes for reading”, but genes that code for a number of proteins which develop phenotypes that aid reading: vision, language, intelligence, memory, etc, etc, etc.

      There are a number of approaches towards separating the influences of genes and shared environment, but the most straightforward is the twin method, which I will describe. Identical twins share an environment and 100% of their DNA (oversimplifying slightly, but the approximation is more than close enough). Non-identical twins also share an environment but only 50% of their genes – they are no more similar, genetically, than full siblings. Therefore we can conclude that if identical twins are, on average, more concordant for a trait than non-identical twins, genetic influences are at play in trait variation. The equations work thusly:


      Genetic contribution to variation = h2
      Shared environmental contribution = C
      Non-shared environmental contribution = E
      identical twin correlation for the trait in question = MZr
      Non-identical twin correlation for trait in question = DZr

      As discussed above, the identical twin correlation arises from genes + the shared environment, so

      MZr = h2 + C

      therefore the leftover “non-shared environment” (includes measurement error)

      E = 1 – MZr

      if identical twins are twice as genetically similar as non-identical twins, then genetic influence is double the difference between the identical & non-identical twin correlations

      h2 = 2(MZr -DZr)

      And this now allows us to solve for C

      C = MZr – h2

      NB: this is not how heritability is actually calculated in the real world: modern-day behaviour geneticists use maximum likelihood model fitting techniques. But essentially the equations presented above allow you to understand what is going on in sufficient detail.

      Another approach is to look at adopted children, who share an environment with their adoptive siblings but no genes (NB, however, that this is not entirely true because even children adopted from very young ages will not have shared a pre-natal environment with their adoptive siblings). Do the IQs of adopted children correlate with the IQs of their adoptive siblings? Or do they correlate instead with those of their biological parents? It is, of course, consistently found that by adulthood the correlation between the IQs of adopted children and their adoptive siblings is zero.

  2. Dylan Wiliam says:

    I wrote about the same issue in an AERA paper in 2005 (for those who are interested, it’s on my website,, under the “Papers” tab). The somewhat counter-intuitive take-away is that, as David says, far from ignoring the genetic contribution to academic outcomes, we should be seeking to maximize it, because when all students have access to the best possible learning environment, the only source of variation is the (currently immutable) genes…

  3. Using a real maths example from our school:
    School fast-tracks three ‘bright’ infants in maths and attributes to ability. The pupils have confident, professional, articulate, numerate parents who have prepared and supported the child in number bonds, times tables, etc from very early on. They receive home tuition during holidays and receive 1:1 support through the difficult learning, which builds confidence.

    Is this rapid progression witnessed by the teacher primarily down to parental genes or parental involvement?

    A teacher sees a child progressing quickly, is unaware of the level of home support, and assumes learning is solely down to the child’s ability. Teachder then reinforces it through positive feedback, expectations, learning goals (halo effect). Is it primarily good genes or the influence of shared involvement at play in the classroom? Is bias in the non-shared environment accounted for in the study?

    Genetic interplay is interesting and has a place in this discussion. It is important to be clear all variables have been isolated and factored into a study in order to reach a conclusion. Do you think this study has?

  4. Fascinating. It makes a lot of sense, too, even at the personal anecdotal level — my maternal grandparents were fluent and avid readers, as was my mother, all from relatively unschooled working class backgrounds. And now so am I and my daughters.
    But that’s by the way. Anecdotal, as I acknowledge.
    What interests me is this para;
    ‘This is deeply counterintuitive. As Asbury puts it, “More school – that is, more environmental input – leads to greater genetic influence rather than greater environmental influence.” (p. 28) As children’s environmental influences become more similar, genetic differences become more noticeable’
    Is that really counter-intuitive? Does it not say – and I may have misunderstood, which is partly why I raise it — that a child genetically predisposed to reading is better equipped than others to pick up and run with the formal reading instruction when it arrives? That presumably sets up a home-school feedback loop. Actually quite intuitive if I have it right. And has implications for the way teachers differentiate their instruction.

  5. MaggieD says:

    Oh, sorry David. In my previous comment I was responding to Jack, not you.

    Hitting the ‘reply’ button doesn’t always seem to get me to where I want to be!

  6. When it comes to the teaching of reading in particular, the field is especially complicated by what people mean by ‘traditional teaching’.

    The chart above shows statistics from North America, Australia and Europe – the first two being English-speaking countries in the main and the latter – Europe – reflecting a variety of languages commonly with a simpler alphabetic code than English.

    Whereas I have no doubt that genetics (inherited traits, tendencies, talents, difficulties) does make a difference to learning capacity, I suspect that when it comes to the teaching of reading in particular – at home and in school, there are probably many more complications than meet the eye – certainly more than are mentioned in this blog posting.

    • David Didau says:

      You’re right Debbie: so many more. That said, I’ve been having a squizz at the study I’ve linked to and they do seem to try to account for and isolate these variables in reading as well as mathematics and PE (interestingly science seems different and genetics plays a significantly smaller part in students’ success) but the whole point is that we can best know instruction is effective if the differences between children can be best explained by genetics. This is good news.

      • ChrisN says:

        “the whole point is that we can best know instruction is effective if the differences between children can be best explained by genetics. ”

        I would be very cautious about a statement like that. Type 1 diabetes is in large part hereditary; before effective treatment, whether or not children died of diabetes was best explained by genetics, but that didn’t mean the treatment was effective.

        Similarly, if a method of teaching reading doesn’t teach decoding well, then children who are genetically predisposed to dyslexia, and thus need such explicit instruction, will score low on measures of reading. So what is needed is instruction targeted appropriately to minimise such genetic factors.

        • David Didau says:

          Everyone needs explicit reading instruction. That’s the point. It’s counter intuitive, but what we need is instruction that maximises genetic differences. See comments above.

          Have you read my blogs on dyslexia?

          • ChrisN says:

            I understood what you meant, but was suggesting care needs to be taken in how it is phrased. The situation we have now in most English speaking countries is one in which it is inadequate reading instruction that is maximising genetic differences (by not paying sufficient attention to the need for systematic phonics instruction, and thus leaving genetically at-risk children to flounder).

            It’s true that if instruction is maximally effective, then the remaining differences between children would be (mostly) genetic. But you can’t reverse that to claim “we can best know instruction is effective if the differences between children can be best explained by genetics.”

          • David Didau says:

            I agree with you about the need for systematic synthetic phonics instruction; the problem is the poor reading instruction minimises genetic differences. Why can’t you reverse the claim?

    • Andrew Sabisky says:

      It’s worth noting that behaviour genetics does not aim to explain all the complicated interactions that form individual processes of reading development, just to give solid estimates of the respective contributions of genes, environments of different kinds, etc. It’s a method of analysing the variations in the outputs of a population of black boxes; it does not analyze what actually goes on inside the black boxes.

      Complexity at one level does not preclude simplicity at another. Let’s say I am getting hit by a taxi – even though all the interactions are incredibly complex at an atomic level, this does not mean that we cannot describe what is going on in simple terms (in fact, it’s clearly preferable to do so).

      Molecular geneticists have a similar reaction – confusion – when quantitive geneticists tell them that genetic contribution to most variation in complex traits is largely additive because they know how complex the interactions are at a lower level. Everyone else has the same reaction when behaviour geneticists say that phenotypic outcomes can largely be represented as a simple P = G + E equation, because we all know how complex the interplay between genes and environments are in our everyday lives, so when we’d expect lots of statistical interactions to show up in twin and adoption data (plenty of GxE as well as/instead of G+E). But then again, think about my example of getting hit by a car, and it all makes a bit more sense.

      • Lucy P says:

        So on a very basic, anecdotal level the identical twin I taught perhaps illustrates your point. Shared genes, shared environment, both premature but different reading ability to his twin. Inside the black box he is different? And personality/character must also play a part? I.e. one perhaps ‘coped’ with the struggle of learning better than the other.

  7. heatherfblog says:

    It is interesting that we see that move to the right of the bell curve in Far Eastern countries with maths.

  8. mrbenney says:

    Another very, very good blog. I think your bell curve comment is absolutely right. Whatever we are interested in measuring/improving, hard work and great teaching can shift that bell curve to the right.

  9. As long as it is noted, however, that whilst there needs of course to be an acknowledgement that all children are ‘different’ both in terms of their inherited differences and in terms of the differences in their environmental experiences, teachers, nevertheless, must not overly attribute apparently weaker learning to ‘within child’ issues only.

    This is particularly the case with the teaching of reading. You will find that people in the field of reading instruction who favour explicit and systematic phonics teaching strongly suggest that good quality teaching and content can seriously work wonders whatever the inheritance of the learners themselves.

    The worry is that people do not fully appreciate this because they ‘turn off’ when it comes to the subject of the debate about reading instruction – hence the level of phonics detractors in England.

    I hope it is not a turn-off, however, to note that there are currently schools achieving almost 100% of their children reaching or exceeding the benchmark in the Year One phonics screening check. I do think that this is relevant to this thread as what I am trying to demonstrate is that ‘regardless’ of genetics and environment, we can be successful in teaching all of the children to read despite our complex alphabetic code.

    • David Didau says:

      Absolutely right – I *think* that’s what everyone on this thread is saying too. I certainly am. In an ideal education system genetic influences on the differences between children would be closer to 100% – that fact that it’s only 80% speaks of the asymmetry of non-shared experiences. And yes, the overwhelming majority of children can learn to read – I’m struck by Stanovich’s observation that there is zero correlation between cognitive ability and the ability to decode.

  10. Presumably the claim is the variation in ability is 80% attributable to heredity. That is a a totally meaningless claim. Ability in the sense the IQ fetishists mean it is 100% in the genes. Knowing that tells us exactly nothing.

  11. […] For whosoever hath, to him shall be given, and he shall have more abundance: but whosoever hath not, from him shall be taken away even that he hath. Matthew, 13:12 The Matthew Effect has become something of a truism. Those with find it easy to acquire more, whereas those without are trapped into a vicious cycle of  […]

  12. pahi smith says:

    you research too good. your point of view is directly points to mutation. and i said that if mutation was not occur ih all kind of living organism than evolution is not possible. I agree that genetically variation is necessary but this is done by their own.

  13. TheOtherDrX says:

    With regard to intellectual ability, the hereditary element is clearly a poygenic trait, that is, influenced by many gene variations contributing to the phenotype for whatever is being measured (Maths, literacy, languages, athleticism, music). Some of these abilities at least are linked. See

    for a recent and well designed study. The results are not all that surprising: Maths and Literacy ability have a inherited polygenic component, and are linked, that is, if you are good at maths, you are more likely to be good at literacy. With regard to heritability of intellectual ability, a ‘best guess’ is that the academic ability of off-spring will be somewhere between that of the two parents on average (as it is polygenic), but we all know examples where that is not the case.

    From a genetics point, this really gets tricky in separating our genetic and environmental effects. Parents who are ‘genetically intellectually more able’ and were given an opportunity to thrive educationally are presumably more likely to give more beneficial additional tuition to their offspring at home. Those parents who were ‘genetically intellectually more able’ who didn’t have this opportunity during their education might not. But the offspring might still do far better than the parents due to the genetic component. Suddenly the blurred lines between genetics and environment, but not necessarily the environment of the child becomes a factor.

    The Matthew effect comes into play when there is selective pressure allowing one phenotype to gain advantage over the other. In the lab, I routinely make bacteria resistant to antibiotics by a simple trick of introducing an antibiotic resistance gene. These bacteria behave the same as normal bacteria, but when you add a sective pressure, the resistant bacteria thrive as they have an advantage. Withdrawing the opportunity to thrive reverts them all to behaving the same (unless they are dead). In education, the more opportunities to thrive/learn, and to progress at the individuals maximum rate will always widen the range of abilities. In theory this should shift the outcomes curve to the right for all… That is unless you ‘kill-off’ (metaphorically, from a learning perspective) all the individuals with lower ability, with the intervention.

    The question is: Should we fight against the Matthew effect, or embrace it? Embracing it by shifting everyone to the right by some environmental/educational intervention is good, right? Err, not necessarily. Some will argue the higher-end shift to a much larger degree than the lower end is education inequality and that’s bad (we will call this ‘relative killing-off’…), even though the lower end are doing ‘better’ than without the intervention. This makes statistical interpretation of outcomes of so many education studies very difficult, and translation into sensible policy near-impossible. As a geneticist (sort-of), I doubt whether genetics will ever play a part in any education policy, but will make a very interesting intellectual area of scietific study for many years to come.

    • Kathryn Asbury says:

      Agreed it is difficult but I am very keen to keep on thinking about this. Moving the distribution to the right – especially if we can identify ways of super-charging the shift of those on the left hand tail of the distribution (thereby closing the gap a little at the same time as raising the mean) seems like something worth fighting for (but I have been called a Pollyanna before today). Maybe non-shared environmental influence is the key, or perhaps a better understanding of genotype-environment interplay? Worth thinking about but I agree there are no easy solutions here. FYI I wrote about the Nature piece you linked to here:

  14. […] For whosoever hath, to him shall be given, and he shall have more abundance: but whosoever hath not, from him shall be taken away even that he hath. Matthew, 13:12 The Matthew Effect has become something of a truism.  […]

  15. […] Many, many children, however, are clearly making it to secondary school without these tricks, and this is another factor that leads to the ‘Matthew Effect’, the tendency of the word rich becoming word richer and word poor becoming word poorer (Didau explains further here). […]

  16. […] David Didau‘s excellent post, Reading Ability: Nature of Nurture? is here. […]

  17. F. Aetius says:

    It seems to me that the more research into genetic effects, the better: the higher the genetic component of education outcomes, the more equal our system is (ie the less it depends on school and societal variability).

    Screw PISA and national attainment data, the best way to judge the effectiveness of our nation’s schools is through research into variability due to genetics.

  18. Amelia Peterson says:

    I’m interested by the argument that: ‘we can best know instruction is effective if the differences between children can be best explained by genetics’. How are we supposed to know when that is the case? If environment-genetics interplay starts at birth (or in womb…) then we have no baseline. Assuming we don’t want to accept the socio-economic verbal skill gap that children start school with (still pretty considerable according to latest study with good data set: what should we do with the information that some of that gap may be down to genetics?

    (Another way to put this: I’m concerned by the neatness of 80% describing the situation for ‘ability’ in general, and the comment above that implies different ‘abilities’ are ‘phenotypes’ of particular genes. I’m not an (epi)geneticist or a neuroscientist, but I gather the situation is rather more complicated (see e.g. Nessa Carey and Kurt Fischer to counter balance Plomin), and given the way brains develop there is scope for environmental influence to go ‘all the way down’. This isn’t just utopian ‘oh I wish we were all the same’ – genes must play a role in how physiology develops – but if in any given case we don’t know what that role is, I don’t see how a summary estimate averaged across individuals and skills provides any actionable information)

  19. Leigh Taylor says:

    Dianne McGuinness’ ‘ Why children can’t read’ should be required reading for the whole teaching profession. She clearly explains how children learn to read, the possible difficulties and how to effectively teach reading to all children ‘ unless they have such a low IQ that their memory and vocabulary is severely impaired’. Her conclusion is that only fidelity to a scientifically evaluated systematic synthetic phonics programme will produce fluent decoders and therefore enable effective comprehension.

    Why in 2015, 10 years after Sir Jim Roses’ report, some teachers still insist on struggling readers ‘learning’ whole word memorisation strategies such as Magical Spelling and Reading or looking for clues in the sentence is unbelievable.

    I hope that the bodies and individuals, such as the EEF, Sutton Trust and Greg Brooks, continue to evaluate systematic synthetic phonics programmes and prove that reading can be only be effectively taught if the code is explicitly explained and practised and that children are not confused by charlatan methods.

  20. […] The assumption that if student cannot read they must be stupid  seems to stack up on the face of it – after all, they can’t do something we find effortless. And this is the problem: the ability to read is so effortless we tend to think of it as somehow natural. It’s easy to see how this happens. After all, the ability to learn spoken language is innate. We receive no formal instruction; we just pick up speech from our environments. Reading though is different. No one, no matter their ability, just learns to read. You have to be taught. You can read more about this here. […]

  21. […] that the rhetoric around ‘closing the gap’ is misguided; the best we can probably do is seek to move the entire bell curve to the right. I also agree that there is good reason to be sceptical about the EEF. She also makes a fair point […]

Constructive feedback is always appreciated

%d bloggers like this: