On the face of it, nothing. But let’s follow the supply chain. Every year the world produces over 60 million tonnes of electronic waste: it’s the fastest-growing waste stream on the planet. A significant portion still ends up recycled informally through processes that release heavy metals into local environments. And new research is finding that those same metals are associated with increased rates of dyslexia in children who’ve been exposed.

The link between pollution and health isn’t new. Nightmares about clowns? Forget it. To live in 2026 is to have nightmares about crisp packets and tap water, along with all of the other general 2026 vibes.

But pollution and reading? Over the last five years, researchers have started finding links between environmental chemical exposures and reading difficulties.

Which raises a question: is the environment shaping how reading develops, and for whom?

TL;DR. Yes.

The heritability gap

You probably already know dyslexia runs in families. If you’re dyslexic, you might not be the first person in your family to know what that’s like. Genetics clearly plays a role, but it’s never been able to fully explain the patterns we see. There’s a persistent gap between how heritable dyslexia looks in families and how much the DNA itself can account for.

What’s starting to fill that gap is the environment, and in how environmental exposures can influence how our genes behave. Not by changing the DNA, but by changing which parts of it are active.

Your genes aren’t a script: they’re more like a set of possibilities, and your environment helps decide which ones get switched on. We share more than genes with our families: we share air, water, diet, stress, housing and a dubious love for Radiohead (thanks dad). If those shared exposures are influencing gene expression, then some of what looks like straightforward inheritance might be something more complicated, and more modifiable, than assumed.

Where your old phone ends up

This brings us back to the phone in your pocket…or more precisely, to where it could have ended up.

In Guiyu, in China’s Guangdong province, over 5,000 small workshops once processed discarded electronics through informal recycling. The Chinese government has since intervened by investing in a regulated industrial park and restricting e-waste imports. The city looks very different now. But heavy metals don’t really listen to policy timelines. They persist in soil and water long after practices change.

Researchers at Shantou University recently screened 2,520 primary school children in the area for dyslexia. They found that children with higher exposure to mixtures of metals (particularly chromium, nickel, and lead) had significantly higher rates of dyslexia. A 2026 systematic review pulled together all available studies on heavy metals and reading difficulties, and the direction was consistent: environmental metal exposure is showing up as a risk factor for dyslexia, across multiple studies and populations.

What makes this relevant beyond Guiyu is what it tells us about how developmental difficulties develop. We’re used to thinking about dyslexia in terms of a child’s phonological processing, their working memory, their hearing, etc. This research also tells us that what’s happening around the child matters too.

For anyone working in education, that’s important. It means the profile of strengths and difficulties a child brings into your classroom has been shaped by more than their neurology. It’s been shaped by their environment in ways the child has no control over, and we need to be sensitive to that.

Closer to home

The e-waste research might feel distant from your daily life, but the environmental story doesn’t stop at recycling sites.

A 2025 study looked at the chemicals in everyday consumer products. Researchers measured levels of common phenol compounds including bisphenol A (which you’ll find in food packaging, water bottles, and receipt paper) in over a thousand children, alongside their genetic susceptibility to dyslexia.

Children with higher BPA levels had a higher likelihood of dyslexia. But the really interesting finding was what happened when the researchers looked at genetics and chemical exposure together. Children with both high genetic susceptibility and high BPA exposure had notably greater odds of dyslexia than children with just one or the other. Neither factor alone told the full story. The combination mattered.

This is what gene-environment interaction looks like. And it’s worth knowing about, not to add BPA to your worry list (though maybe stop microwaving your dinner in it), but because this research is shifting our understanding of what dyslexia actually is. That shift matters not just how we identify, support, and think about children who struggle with reading, but when.

Stress as a biology

There’s a third environmental thread. We tend to think of stress as psychological and something that you feel or cope with. But chronic stress, particularly in early childhood, changes the body’s chemistry, and over time, it can change how genes are expressed.

Reading is an evolutionary newcomer. There’s no gene for it. The brain repurposes circuitry that evolved for other things, and that repurposing depends on systems that also regulate our stress response. In a typically developing child, moderate stress is part of how the brain calibrates itself. But when early-life stress is chronic and overwhelming, those regulatory systems adapt and reading fluency (which the brain never specifically evolved for in the first place) can become collateral.

Researchers have identified specific epigenetic changes associated with this adaptation. Changes to how certain genes involved in stress regulation behave, which are protective in a high-stress environment but come at a cost to the kind of learning that reading requires. The brain is making a trade-off.

The stressors that drive these adaptations aren’t randomly distributed. They tend to cluster with poverty, with pollution, with housing instability, and with educational disadvantage. Not always of course, but the children most likely to be carrying the biological signature of chronic stress may also be more likely to be the same children most exposed to the environmental chemicals we’ve been talking about. The pathways overlap and so the exposures compound. This isn’t a problem individual families created and it isn’t one they can solve alone. It’s regulatory. It’s systemic. It’s about which communities get protected, and which don’t.

So what does this actually mean?

If you’ve made it this far without panic-buying a glass lunchbox, well done.

This research is not saying dyslexia is caused by pollution, or by stress, or by your old phone. Dyslexia is a form of neurocognitive variation that exists across populations, cultures, and environments. It always has. It’s just now more of a thing since, well, reading print became more of a thing. And like all neurocognitive variation, it’s a whole way of being in the world, not just a relationship with text. This research doesn’t change that.

What it does is make the picture more complicated. Sorry. I know that’s not quite what we need right now. The environment interacts with genetics in ways that shape how reading develops, and we’re only just starting to understand what that means.

The researchers themselves are pretty concrete about what should happen next. Screen earlier, screen wider. In high-pollution areas, consider environmental exposure as part of the picture, not just cognitive assessment. Regulate the chemicals that are showing up in the data. Investigate what buffers against them.

Some of this research was only published in the last few months. One study lands in April. So far (to my knowledge) the questions have mostly been asked in China. But every country has its own environmental history, and the UK is absolutely no exception. Whether any of that relates to how reading develops here is a question nobody has asked yet.

References

Doust, C., Fontanillas, P., Eising, E. et al. Discovery of 42 genome-wide significant loci associated with dyslexia. Nat Genet 54, 1621–1629 (2022). https://doi.org/10.1038/s41588-022-01192-y

Ebadi, M., Narimani, M., Aghajani, S., Eyni, S., & Fattah Zade Ardalani, G. (2026). The Role of Heavy Metals as a Biomarker in the Pathogenesis of Dyslexia and Specific Learning Disabilities: A Systematic Review. Neuropsychobiology, 85(1), 50–58. https://doi.org/10.1159/000548323

Gialluisi, A., Andlauer, T.F.M., Mirza-Schreiber, N. et al. Genome-wide association study reveals new insights into the heritability and genetic correlates of developmental dyslexia. Mol Psychiatry 26, 3004–3017 (2021). https://doi.org/10.1038/s41380-020-00898-x

Kershner J. R. (2021). An Evolutionary Perspective of Dyslexia, Stress, and Brain Network Homeostasis. Frontiers in human neuroscience, 14, 575546. https://doi.org/10.3389/fnhum.2020.575546

Kershner J. R. (2024). Early life stress, literacy and dyslexia: an evolutionary perspective. Brain structure & function, 229(4), 809–822. https://doi.org/10.1007/s00429-024-02766-8

Xiang, Z., Wang, H., Zhu, K., Liu, R., Zhao, S., Fan, H., Chen, Q., Zhu, B., & Song, R. (2025). Phenol exposure, polygenic risk score, and dyslexia in Chinese children: Gene-environment interaction. Environmental pollution, 379, 126536. https://doi.org/10.1016/j.envpol.2025.126536

Yu, X., Zhang, X., Wen, W., Lin, X., Xia, X., Wang, D., Wu, K., & Huang, Y. (2026). Childhood dyslexia risk elevated by heavy metal mixtures from e-waste: A machine learning–driven mixture modeling study. Environmental Pollution, 394, 127745. https://doi.org/10.1016/j.envpol.2026.127745

Your eyes are not moving smoothly. They are jumping. Short rapid movements called saccades. Landing, skipping forward, sometimes back. Your brain is predicting what's coming next and guiding where attention lands.

You’ve probably seen those decades old memes where a sentence contains the the same word twice, and nobody notices. The brain is a prediction machine, hasn’t found anything incongruous and so fills in and moves on. I haven’t done that to you just now (honestly…or have I?) but you were ready for it and are now likely going back and checking.

Within all of this, in a span of milliseconds that you will never consciously experience, a word becomes a word. For you, right now, that process is so fast and so automatic it likely produces no sensation at all.

If that is what just happened to you as you read this, then you are an entirely typical reader.

Congratulations.

As a neurotypical person, I recognise that vantage point is dangerous.

In the UK, we are currently in a National Year of Reading. On dyslexia websites, on GOV.UK, at parents’ evenings, assembly slides, you may come across this statement:

“Reading for enjoyment is the single most important indicator of a child’s future success.”

I’m not entirely sure what it’s meant to do. Inspire, perhaps. Signal priority. Get children reading.

The voices below are taken from peer-reviewed research.

Leo has dyslexia and is in primary school.

“There's a big difference in what I do at school than all the other kids, it's like me wanting to be like really good at something and then like everyone else in the class is like really good at it except for me, and it's just like well what's the point of me trying that if I can't do it.”

Maisie, too.

“It just feels weird, it just feels weird and sad because I don't know what's happening in school…it's bad because all the other kids know what's happening or like they can do it properly, but I can't do it because I have dyslexia yeah and it's just really annoying.”

Also Arlo.

“You can get stressed a lot of the time and then usually get angry if like after school cos you've had like a hard day…when I get home I'm stressed, it gets like if I'm doing an assignment I get really angry and it gets really difficult…like if it's like really [emphasis] difficult…I sometimes cry and get really emotional.”

Tom is in secondary school and has dyslexia. His class is reading aloud with each child taking a paragraph, one after another, working around the room. Tom can see how many children are between him and his turn. He is counting.

“Each student would stand up and read and then it would go on to the next person for a paragraph each and like as soon as it's coming around to me I'm just like “oh, no”, like everything is going on in my mind and I'm like “oh no this isn't going to be good”, but I just have to get on and do it because there isn't anything else to do.”

Tom is describing a stress response. The anticipation of uncontrollable exposure triggers the flip from reflective to reflexive brain function. Chemically, stress floods prefrontal networks with catecholamines. Attention, working memory, and flexible thinking blanks.

How can Tom find pleasure for reading here?

As well as impairing learning in the moment, chronic stress causes lasting architectural changes in those networks. The brain adapts to surviving the classroom.

It really does not have much left over for enjoying a book.

Issy has dyslexia. Her school is trying to help.

"I don't really like teachers making it obvious that I need to take a test out of class and it's really annoying… and then also like I got this C-pen thing and it always like gives me too much attention, like, I don't really like having that much attention on me."

The C-pen was provided to help but when it was used, it announced her in a way she had not chosen to be visible.

Research across dyslexic students’ school years finds that support which marks you out can fracture the identity it is trying to protect (Lithari, 2019, 2023). Children who did not receive meaningful support, or whose support felt unhelpful, visible and embarrassing, had greater difficulty constructing a stable academic identity. Those who received support that felt appropriate to them did not. The difference was whether the child had been included in deciding what that support looked like.

Sometimes, it really can be that simple.

Most dyslexia intervention is built around the cognitive. Phonological processing, decoding, fluency. Adaptive strategies. These matter. But a child’s sense of themselves as a learner shapes outcomes too and it is just as important that is taken into account.

I see your picture of Albert Einstein. Seventy years old, tongue out, captioned with the reassurance that he had dyslexia too. What eight-year-old child, sitting in a classroom right now, struggling to decode the words on the page in front of them, is supposed to find themselves in that?

We always mean well.

But meaning well sometimes means asking uncomfortable questions.

In a National Year of Reading, how is dyslexia celebrated?

We have built reading cultures from the neurotypical vantage point, and perhaps, we have celebrated dyslexia from it too. The poster, the challenge, the famous face on the wall (if not Albert Einstein…Emma Watson?).

The assumption is the same either way.

Reading for pleasure was never really about reading.

It is about narrative. Other minds, other lives, other worlds. The thing that happens when a story gets inside you and changes how you see.

Print is the delivery mechanism which is typically valued. And somewhere along the way we confused the vessel for the thing itself.

A child listening to an audiobook with total absorption is doing exactly what we want reading for pleasure to do. A child performatively completing a reading challenge without engagement is not. And yet we measure the first child as not reading for pleasure, and the second as succeeding.

So evaluate your programmes. Not just who is reading, but who is struggling to performatively demonstrate it. Who hasn’t received a certificate or accrued enough reading stamps. Who cannot tell you what they last loved, in any format, because the question of enjoyment stopped feeling like it applied to them.

Because that child is there in every classroom. And their voices tell us exactly what it means to them to sit in a reading culture that was not designed with them in mind.

And then we tell them, and others, and ourselves…

“Reading for enjoyment is the single most important indicator of a child’s future success.”

They read it. They understood it perfectly.

That is precisely the problem.

Please read this paper. It’s open access I believe:

Wilmot, A., Pizzey, H., Leitão, S., Hasking, P., & Boyes, M. (2023). Growing up with dyslexia: Child and parent perspectives on school struggles, self-esteem, and mental health. Dyslexia, 29(1), 40–54. https://doi.org/10.1002/dys.1729

References

Arnsten A. F. (2015). Stress weakens prefrontal networks: molecular insults to higher cognition. Nature neuroscience, 18(10), 1376–1385. https://doi.org/10.1038/nn.4087

Lithari, E. (2019). Fractured academic identities: dyslexia, secondary education, self-esteem and school experiences. International Journal of Inclusive Education, 23(3), 280–296. https://doi.org/10.1080/13603116.2018.1433242

Lithari, E. (2023). Academic identity development: school experiences and the dyslexic learner. International Journal of Inclusive Education, 27(8), 851–867. https://doi.org/10.1080/13603116.2021.1879947

Lithari, E. (2025). Dyslexic students and their early experiences of support in the first years of secondary education in England. Education 3-13, 53(6), 922–936. https://doi.org/10.1080/03004279.2023.2247403

Wilmot, A., Pizzey, H., Leitão, S., Hasking, P., & Boyes, M. (2023). Growing up with dyslexia: Child and parent perspectives on school struggles, self-esteem, and mental health. Dyslexia (Chichester, England), 29(1), 40–54. https://doi.org/10.1002/dys.1729

A doctor appears with a consent form for a new procedure.

It might help, they say.

The research is preliminary. The mechanism isn’t fully understood. It hasn’t yet been tested on people like you, your demographic, your profile. But the efficiency gains look promising, and a trial needs participants.

You ask what “help” means. Compared to what? Measured how? And what happens if it doesn’t work?

You would not proceed without answers.

The announcement

The government has announced today that AI tutoring could benefit 450,000 disadvantaged pupils.

The tools will “adapt to individual pupils’ needs”, “provide extra help when they get stuck” and “identify where they need more practice to master their lessons”. The goal is for pupils to “catch up with their peers”.

This sounds uncontroversial…Benevolent, even. Extra help, tailored support, closing gaps. Especially for those who are disadvantaged. Who would object to that?

But this does raise an important question we should be asking ourselves.

Do we want technology that adapts to children, or children who must adapt to the technology?

The answer is to that is determined by design.

The announcement promises tools will be “co-created with teachers” and “robustly tested” for safety. But once the underlying model of learning is fixed, co-creation and testing operate within it, not on it.

This is not a new concern. For years, the AI in Education research community has documented how systems trained on typical learning trajectories handle difference not by adapting, but by normalising it away. As Kaska Porayska‑Pomsta (2024) cautions, when learning diverges from what a system is built to recognise, the system does not adapt, it stabilises itself. Who is the cost of that stability borne by? The learner, particularly where learning is least predictable.

The system does not shape to the learner.

The learner is shaped to fit the system.

Those warnings do not appear here. The announcement assumes the questions have been answered. That we know what learning is, what progress means, what support serves it. That scale and efficiency are the remaining problems, not epistemology or ethics.

“Help” is not a neutral category. Get those assumptions of support wrong (or worse, leave them unexamined to hit the six month deadline you’ve set yourself) and you don’t get efficiency at scale. You get the wrong thing, delivered consistently, to those with the most at stake with the least capacity to object.

What teaching requires

A child says “twelve.”

Did they count? Did they remember? Did they guess? Are they consolidating something solid or repeating something half-understood? The answer itself tells you almost nothing. What matters is what is hidden.

This is what teaching is: trying to see what you cannot directly observe going on inside the head of a four-year-old. You watch for hesitation. You notice tone. Gestures. Expression. You remember what worked last week and what didn’t. You’re likely wrong constantly. But you can sense when you’re wrong, and that doubt itself becomes information. You check. You adapt.

What AI cannot do

AI systems also interpret learners, but they tend to do so through what can be measured instantly: right answers, wrong answers, response times, error repetition. From this they build a model and act on it. If the model is uncertain, that uncertainty is resolved algorithmically. The system picks a path and continues.

It cannot do what human interpretation does: stop because something feels off. Question its own reading. Wait because maybe the struggle is necessary. Revise its understanding. These moves depend on human judgement, and on accepting that you might be wrong about what is happening in order to adapt.

And this is the problem. What matters most may resist simple measurement.

Some of the most important features of learning do not register as progress in the moment. Performance can dip before it improves. Understanding may emerge through confusion. Agency develops by learning to stay with difficulty, not by having it removed.

Systems designed to optimise measurable performance (without a human in the loop) don’t really have way to distinguish these processes from error. What they cannot represent, they resolve. In doing so, they stabilise performance by smoothing away forms of struggle that are not obstacles to learning, but part of it.

AI tutoring may improve performance. But what kind of performance? And is that performance what a human would recognise, simply, as learning?

The uneven distribution

And for whom does this definition of learning become acceptable for?

Because educational inequality, really, is not only about access. It is about what becomes acceptable for different children. When we decide that automated support is an acceptable substitute for human judgement for disadvantaged children, we aren’t really closing gaps. We’re re-drawing the baseline. We are saying that some children are entitled to human attention as standard, whilst others are expected to make do with systems designed to approximate it.

The problem truly is not the presence of technology, but the way its limits are assigned to particular children. When a tool encodes a particular philosophy of learning, it also determines who is expected to live with what it cannot do. And when those limits concern judgement, interpretation, and care, the work at the heart of learning, well, the consequences are not shared evenly.

What the evidence can and cannot tell us

It is tempting, when responding to announcements like this, to divide the world neatly into optimism and scepticism. Either technology will transform learning, or, at best, it will disappoint. Get with the times, or be a luddite.

Decades of research on educational apps suggest that digital tools can support learning. But they do so conditionally, unevenly, and within bounds.

When positive effects are found, they tend to be tightly coupled to the specific skills the app targets. Practice improves performance on practiced tasks. Number games strengthen number skills. Pattern drills improve speed and accuracy. This is not trivial, and for some learners it can be genuinely helpful. But it is largely a story of near transfer: improvement that stays close to the original activity.

Evidence for broader generalisation is thinner. Gains in conceptual understanding, problem-solving, or long-term retention are less consistently observed, and when they do appear, they are often sensitive to design choices that are easy to overlook in policy discourse. Explanatory feedback matters more than praise. Structured progression matters more than surface personalisation. Integration into classroom routines matters more than the app itself.

In other words, how the technology is designed and used matters at least as much as whether it is used.

This should give pause to claims framed primarily in terms of reach. Scale does not amplify all mechanisms equally. What scales most reliably are forms of practice and correction. What scales poorly are judgement, interpretation, and responsiveness to context.

There is also the question of time. Much of the evidence for educational apps relies on outcomes measured immediately after an intervention. Fewer studies examine whether gains persist months or years later, and those that do report mixed results. Some effects endure. Others fade. Some disappear entirely once the novelty or structure of the intervention is removed.

This matters because learning is not simply what is visible at the end of a session. It is about what remains available to the learner when support is withdrawn.

Claims about benefit for disadvantaged pupils require even greater care. While there is some evidence that learners who are struggling can show substantial short-term gains with well-designed apps, the methods used to identify “underachievement” are often crude, and effects can be inflated by statistical artefacts rather than genuine change. More importantly, there is little evidence that such tools reliably reshape a learner’s long-term relationship with learning itself.

The evidence, taken as a whole, does not condemn educational technology. But neither does it justify the confidence often projected onto it.

What it suggests instead is something more modest and more demanding: that digital tutoring tools can support learning when they are narrowly targeted, carefully designed, thoughtfully implemented, and held within a broader educational ecology that does not confuse improvement with understanding.

It reminds us that learning technologies do not fail because they do nothing. They fail because they do something very well, and we mistake that something for everything.

Who is being consulted?

The research on this exists. It is decades deep. It is substantial.

The AI in Education research community has documented what these tools can and cannot do, under what conditions they help, and where the risks lie. That knowledge is readily accessible. It should inform policy.

The government has set a six-month deadline for co-creation with teachers. The question is whether it will also consult the researchers who have spent careers understanding how learning works, what interpretation requires, and what happens when systems designed for efficiency encounter the complexity of actual children.

When systems are introduced at scale for 450,000 children, the evidence base should be proportionate. Short-term gains on narrow measures are not the same as long-term support for learning. Promising efficiency is not the same as demonstrating benefit. And when interventions are tested primarily on disadvantaged children, someone should be asking whether that is equity or expedience.

The tools will be built. The pilots will run. What remains uncertain is whether the government will draw on the expertise that exists to shape what gets built, or proceed as if decades of research do not apply.

The traces left will remain long after the software is created, the pilot ends, and the Education Secretary moves on.

And those traces leave permanence.

References

Department for Education (2026). 450,000 disadvantaged pupils could benefit from AI tutoring tools. GOV.UK.

Porayska-Pomsta, K. (2024). From algorithm worship to the art of human learning.

Porayska-Pomsta, K. A Manifesto for a Pro-Actively Responsible AI in Education. Int J Artif Intell Educ 34, 73–83 (2024). https://doi.org/10.1007/s40593-023-00346-1

Outhwaite, L. A., Early, E., Herodotou, C., & Van Herwegen, J. (2022). Can maths apps add value to young children’s learning? A systematic review and content analysis. London: Nuffield Foundation.

Dog.

Cat.

Apple.

Pear.

Developmental theory.

Piaget’s stages of cognitive development of course!

A reflex almost automatic in education. Development means Piaget. Piaget means stages. Stages mean something teachers were once made to memorise and then learned, quite reasonably, to ignore.

If this is where your eyes usually glaze over, stay with me: this isn’t a post about Piaget.

This post is about an argument that education is still having, even though it thinks it already won in it the 1990s.

Development Is Not Pedagogy

Now Jean Piaget was an extraordinary scientist who cared deeply about children, and about molluscs. Fortunately for education, he devoted his life to the former. He changed how psychology understands children’s thinking and his influence is still felt across the field.

Of course there are critiques of a theory of cognitive development which was first published in 1936. Piaget acknowledged many of them himself. And of course the theory does not tell teachers how much practice to give, what to present next, or how long learning should take. Stage theory was never meant to function as pedagogy.

Rather than prompting education to refine its understanding of development, the critique froze it. Development was reduced to stage theory, and when that theory did not offer instructional guidance, it became easy to conclude that development itself was beside the point: the argument “I teach children, not stages”.

Developmental science moved on decades ago. Education, broadly, has not.

We are still citing a 1992 interview with Siegfried Engelmann as though it represents the current state of the field. We are still using a Victorian-born scientist as shorthand for an entire discipline. We are still treating “developmental theory” as though it stopped with Piaget, and then dismissing it for limitations that were identified before most current teachers not just entered the profession, but were born. Or their parents born.

This is intellectual complacency dressed up as rigour.

And again, child development is not pedagogy. It was never supposed to be.

This is a category error, and education keeps making it. Developmental theory describes how cognitive, emotional, and biological systems emerge, re-organise, and stabilise over time. It explains how experience interacts with maturation, why the same input produces different outcomes at different points, and what learning costs different children. It is concerned with trajectories, constraints, variability, and developmental cost.

It does not prescribe instructional moves. It does not sequence lessons. It does not tell teachers what to do next. When education demands that it should, and then dismisses it for failing, that failure reflects misuse of the theory…a flaw in the field, not the theory.

In other fields, this distinction is routine. Pharmacology explains how drugs interact with biological systems. It does not tell doctors which drug to prescribe to which patient. That remains a matter of clinical judgement, informed by pharmacology but not mechanically derived from it.

A doctor does not dismiss pharmacology because it fails to provide a prescribing algorithm. They use it to understand why a medication worked, why it didn’t, why one patient needed a different dose, or why another experienced adverse effects. Clinical practice sits on top of that explanatory knowledge. It is not threatened by it.

Education typically has struggled to make the same separation, and then used that struggle as permission to ignore development altogether.

What Counts as Research-Informed

We say we are research-informed. We use the phrase easily & proudly. But if that is the case, then why does education (in the broadest sense) keep reaching for the same arguments to step away from developmental science?

Part of the answer is that not all research serves the same function. Some research is designed to guide action. It tells us what to do next, how to optimise outcomes, how to improve performance under particular conditions. This kind of work integrates smoothly into education because it supports decisions made at scale in fast-paced complex environments, consistent implementation, and demonstrable impact within high-stakes accountability systems.

Other research serves a different purpose. It explains how learning systems change over time. It focuses on variability, timing, and constraint, and on why the same input can produce different outcomes across learners and contexts. This work does not resolve into programmes or checklists. Its contribution is explanatory: it helps us understand what learning demands, why progress is uneven, and where systems are noisier.

The difficulty is not that education values the first kind of research more. It is that the second kind rarely shapes design. If explanatory research is treated as background rather than as a basis for decision-making, schooling will continue to operate as if learning systems were stable, uniform, and predictable.

Nobody who works with children would describe them using the words stable, uniform, and predictable.

And at this point, disengagement from developmental science is a structural convenience.

What We Get Wrong

And that is where the gap opens between calling ourselves research-informed, and engaging fully with the research that explains why learning actually looks the way it does in schools.

That discomfort is often where the conversation stops.

If we ignore this body of research, we make predictable mistakes.

We misread learning failure. When learning is fragile, uneven, or is impacted under pressure, we assume something has gone wrong with instruction, effort, or motivation. Developmental research shows that learning systems re-organise over time, and success can precede stability. What looks like failure is often instability surfacing when demands increase. Doubling down on delivery doesn’t fix that. It could make it worse.

We mistake variability for noise. Much of the work on neurodevelopmental difference, comorbidity, and gene–environment interaction shows that there is no single developmental pathway that most children follow. Variation is not a deviation from the norm; it is the norm. When education treats this literature as specialist or exceptional, it designs systems around an “average learner” who does not exist, and then acts surprised when strain accumulates.

We overestimate what short-term performance tells us. Developmental and connectionist models make clear that performance and learning are not the same thing. Systems can produce correct answers without durable re-organisation. Early gains can mask fragility; later breakdown can reflect earlier trade-offs. When education rewards speed, accuracy, and early attainment as proxies for learning, it optimises for what is easiest to measure rather than what is most stable.

We misunderstand behaviour. Research on stress, adversity, and neurodevelopment shows consistent effects on attention, regulation, and salience. Behaviour changes not because children suddenly choose differently, but because capacity has shifted. When schools respond only through behaviour systems, they manage symptoms while leaving the underlying developmental load untouched.

We displace cost into pastoral systems. Learning draws on regulatory, emotional, and cognitive resources. When academic demands exceed what systems can sustain, the cost does not disappear. It surfaces as fatigue, anxiety, disengagement, dysregulation. Treating pastoral need as separate from curriculum design, pacing, and assessment is an error. These are not parallel problems. They are deeply intertwined, as my other blogs have spoken to.

And finally, we keep recycling old arguments and calling it rigour. By continuing to argue against a long-abandoned version of development, the field gives itself permission to ignore what developmental science actually studies now.

If we claim to be research-informed while sidestepping this work, the consequences are not abstract, or neutral. We misdiagnose difficulty, build brittle systems, and shift the cost of all of this onto the children least able to absorb it.

It is a failure of understanding.

And it is entirely avoidable.

References

Dinkler, L., Lundström, S., Gajwani, R., Lichtenstein, P., Gillberg, C., & Minnis, H. (2017). Maltreatment-associated neurodevelopmental disorders: a co-twin control analysis. Journal of child psychology and psychiatry, and allied disciplines, 58(6), 691–701. https://doi.org/10.1111/jcpp.12682

Hoogman, M., Bralten, J., Hibar, D. P., Mennes, M., Zwiers, M. P., Schweren, L. S. J., van Hulzen, K. J. E., Medland, S. E., Shumskaya, E., Jahanshad, N., Zeeuw, P., Szekely, E., Sudre, G., Wolfers, T., Onnink, A. M. H., Dammers, J. T., Mostert, J. C., Vives-Gilabert, Y., Kohls, G., Oberwelland, E., … Franke, B. (2017). Subcortical brain volume differences in participants with attention deficit hyperactivity disorder in children and adults: a cross-sectional mega-analysis. The lancet. Psychiatry, 4(4), 310–319. https://doi.org/10.1016/S2215-0366(17)30049-4

Karmiloff-Smith A. (1998). Development itself is the key to understanding developmental disorders. Trends in cognitive sciences, 2(10), 389–398. https://doi.org/10.1016/s1364-6613(98)01230-3

Johnson, M. H. (2011). Interactive specialisation: A domain-general framework for human functional brain development. Developmental Cognitive Neuroscience, 1(1), 7–21.

Johnson, M. Functional brain development in humans. Nat Rev Neurosci 2, 475–483 (2001). https://doi.org/10.1038/35081509

Heward, W. L., Kimball, J. W., Heckaman, K. A., & Dunne, J. D. (2021). In His Own Words: Siegfried “Zig” Engelmann Talks about What’s Wrong with Education and How to Fix It. Behavior analysis in practice, 14(3), 766–774. https://doi.org/10.1007/s40617-021-00636-x

McLaughlin, K. A., Sheridan, M. A., & Nelson, C. A. (2017). Neglect as a Violation of Species-Expectant Experience: Neurodevelopmental Consequences. Biological psychiatry, 82(7), 462–471. https://doi.org/10.1016/j.biopsych.2017.02.1096

Sirois, S., Spratling, M., Thomas, M. S., Westermann, G., Mareschal, D., & Johnson, M. H. (2008). Précis of neuroconstructivism: how the brain constructs cognition. The Behavioral and brain sciences, 31(3), 321–356. https://doi.org/10.1017/S0140525X0800407X

Thapar, A., Cooper, M., & Rutter, M. (2017). Neurodevelopmental disorders. The lancet. Psychiatry, 4(4), 339–346. https://doi.org/10.1016/S2215-0366(16)30376-5

Thomas, M., & Karmiloff-Smith, A. (2002). Are developmental disorders like cases of adult brain damage? Implications from connectionist modelling. The Behavioral and brain sciences, 25(6), 727–787. https://doi.org/10.1017/s0140525x02000134

We don’t see things as they are; we see them as we are.

Anaïs Nin

In every classroom there are children who give up at the first sign of difficulty, who shrink at the edge of a teacher’s tone, or who mutter “I’m stupid” when they get something wrong. To an outsider, these reactions can look out of proportion. They’re not. Each is a reasonable response, traced from the child’s own memory.

These pupils don’t need us to deny their feelings or rush to put them right; they need us to notice what those feelings are telling us. They are doing what all of us do: making sense of the present through the patterns laid down in the past.

It is about what they have learned to expect.

Subscribe now

How Priors Shape Predictions

Your brain is constantly processing more than you could ever be aware of. Every flicker of expression, every shift in tone, every background sound is being fed through neural networks that never stop working. This isn’t done to record the present but to predict the future.

This is how the brain keeps you alive: by running forecasts and using what it has learned before to anticipate what might come next. And this is a survival mechanism which has been refined over millennia: it’s better to guess danger and be wrong than to miss it and, well…be dead.

But prediction doesn’t start from nowhere. It rests on priors which are the histories that each and every one of us carry.

Consider something as simple as a stomach ache. One child, drawing from a history of regular meals and gentle care, might interpret the sensation as hunger, or even excitement about a play date with a friend after school. Another child, whose priors might include missed meals, or anxiety, or stress, might predict danger. The physical sensation is identical; the predicted meaning (and therefore the lived experience) is entirely different.

In classrooms, this means that each and every individual pupil can encounter the same signal and experience completely different realities.

One child’s priors might predict encouragement and opportunity; another’s might forecast humiliation and failure, and completely hijack attention.

Same stimulus, completely different worlds.

When Experience Points Toward Threat

For some pupils, their histories have, either consciously or subconsciously, directed their predictions toward caution. A child who has learned that adult attention often means trouble will read a teacher’s approach very differently from one who has learned that adults bring help and comfort. A pupil whose errors have been met with shame will forecast very different outcomes for speaking up than one whose mistakes have been received with patience.

Classrooms are full of signals that need interpreting. For many children, these signals are background noise. For others, they are loaded with risk. What looks like a neutral learning environment to some becomes, for others, a landscape scattered with potential threats.

And the consequences follow a familiar pattern. The amygdala primes vigilance, working memory narrows, attention shifts from the lesson to safety. Energy that could go into learning is instead spent on self-protection. This is absolutely not defiance or deficit: it’s the brain working exactly as it was built to work, but with a different database of prior experiences informing its predictions.

Relationships as New Data

Relationships are where priors shift. Every interaction between teacher and pupil is not just a moment of behaviour management or encouragement, but a data point the brain stores and uses to predict what will happen next. A sharp word, a steady tone, a patient pause: each tells the prediction machine what kind of world it inhabits.

For children carrying priors of shame or exclusion, relationships offer a chance for recalibration. To be corrected without humiliation, to receive attention without fear are experiences that accumulate, slowly teaching the brain that safety is possible.

This is also why equity work, and anti-racism cannot be treated as separate from teaching. Experiences of racism, whether overt or subtle, set powerful priors about belonging and safety. Anti-racist practice does not sit to the side of pedagogy: it shapes the daily signals that tell pupils whether to expect harm or respect, exclusion or belonging.

Relationships, then, are not ancillary to learning. They are the medium through which prediction updates, the ground on which trust and attention can grow.

How Teaching Reframes Expectations

And here’s where hope lives: every classroom interaction provides new experiences that children use to test and update their predictions. Schools and teachers who understand this can design environments that give clearer, safer signals. Evidence strong enough to start shifting what those prediction machines expect.

Predictable openings reduce environmental noise, allowing new data to override automatic caution. When pupils know exactly what to expect in the first five minutes of a lesson, they can focus fully on the learning. It’s predictable.

Transparent progression means pupils stop using cognitive load on trying to second-guess what’s coming next. When learning sequences are consistent, visible and logical, children can focus on the content rather than risk-assessment.

Supporting reframing of mental models/mistakes explicitly reframe mistakes as safe rather than dangerous. When pupils know precisely what will happen when they get something slightly off (and that response is consistently supportive and understanding) their priors about risk begin to shift.

Responsive teaching means the classroom isn’t a one-way street where pupils purely adjust to the teacher’s rules. It becomes a two-way process: the teacher also adjusts to the child. That reciprocity signals safety and fairness, and can start to rewrite old priors.

The Architecture of Safety

It’s important to point out here that predictability alone isn’t enough.

What shifts a child’s outlook is not simply knowing the routine, but encountering experiences that gently contradict and shift old expectations. It is these moments of difference that give them reason to trust anew. Each time a pupil expects danger and finds kindness instead, or prepares for humiliation and meets patience and support, they are given new evidence about what the world can be.

Schools, then, should not imagine that safety comes from controlling or micromanaging every single variable through policy or design. What matters more is creating the conditions, and curricula, where steady signals can be offered and sustained. Over time, these moments weave into a pattern strong enough to loosen vigilance and make space for learning. At its heart, this is an act of care.

And teachers are human too. No classroom can be flawless, and no adult perfectly steady. But what matters is the overall pattern: whether children experience enough, consistently, to begin expecting something different.

These moments may seem small, even ordinary, but they are the building blocks of trust. Slowly and with difficulty, they allow a child to carry a little less fear and a little more freedom. In that shift, they begin to see things not only as they were, but as they can be.

Termly analyses in many schools often point to the same pattern: vulnerable groups - however described - continue to show gaps. And these are often responded to with the same institutional muscle memory: identify, intervene, monitor. Then there is the act of surprise when the gaps remain stubbornly in place.

You might think that after decades of this cycle, we’d question whether we're solving the right problem.

We see a gap and with the best of intentions, think fix, fix, fix, reaching for additional resources, targeted support, bespoke interventions.

But what if this reflex is wrong? What if these gaps reveal less about the individual children, and more about the systems we’ve built around them? What if we've been looking in entirely the wrong direction?

Subscribe now

Learning is much harder than we think

Evolutionary speaking, learning is not one of the brain’s priorities. The brain’s fundamental aim is to try very hard to keep us alive, followed closely by the ability to move, sense, feel things, deal with other people, think, and then - hurrah, learn. Learning is actually really difficult. It requires, among other things, sustained attention, cognitive flexibility, and tolerance of uncertainty. These conditions aren't necessarily equally available to all children…and that’s not necessarily because of differences in ability, but because of differences in the cognitive and emotional resources they bring to each and every moment of their learning.

And it goes without saying that behind every 'data point' is a real child with a real story. A pupil worrying about where they'll sleep over the weekend carries that into every lesson. A child who heard shouting at home this morning is using energy to feel safe that others can put toward learning. A pupil processing trauma - whether recent or ongoing - is doing the extraordinary work of getting through the day whilst also trying their hardest to succeed academically.

These are human experiences that deserve our deepest respect. These children are showing remarkable strength simply by walking through our doors each day. They deserve teaching that honours what they're carrying, understands what they're managing, and responds with both professional skill and genuine care.

And these cohort-level gaps? They don’t reflect the limitations of these children, but systemic blind spots.

Reading the signal differently

This then transforms how we understand vulnerable group data. Instead of reading gaps as evidence of pupil need, we can read them as honest feedback about whether our institutions are working for everyone. The data becomes a way of listening: showing us where our pupils are finding it hard to connect with learning, where our communication and practice may not be aligned, and where the environment we've created may not feel safe or accessible enough for some children.

The question shifts from "which children need more support?" to "what are we doing that makes learning (already difficult) harder than it needs to be?"

The cohort-level data stops being a list of problems and becomes a signal about our practice.

The deeper implications

Taking that signal seriously changes everything about how we approach equity. This truly isn't a new concept I’m writing about here, but this perspective suggests that equity isn't something we add to education through targeted interventions. It's something that emerges when we design systems that work for all learners from the start.

The experiences of pupils facing additional challenges offer important insights about how our systems really work. Their outcomes reveal the hidden assumptions embedded in our practices: this is about learning from their experiences to improve practice to serve our most vulnerable children, and in-turn all children, more & most effectively.

The sequence that matters

This reframing doesn't diminish the importance of individual support: it clarifies when that support is most needed and most effective. The question becomes one of order: we can't distinguish between what pupils genuinely need as individuals and what they need because our systems aren't yet inclusive enough.

Once the causation is clear, individual interventions become more targeted and successful. Instead of asking pupils to jump systemic barriers whilst also meeting their own needs, we remove the hurdles first. Only then can we see where additional support truly makes the difference, rather than where it is compensating for inadequate design.

From signal to practice

If vulnerable group outcomes are signals about our systems, then leadership becomes about interpretation. But what does that interpretation look like in practice?

The first shift is diagnostic. Instead of asking which children need more support, we ask where our systems create unnecessary resistance. This isn’t semantic wordplay. It changes what we examine:

• When gaps appear, do we examine/blame what our pupils lack, or what conditions we've failed to create? And where?

• Do our measures reveal extraordinary pupil resilience, or do they expose how our assessments are designed?

• When pupils with diagnoses struggle, do we read it as part & parcel, or as information about how our systems work for pupils with their needs?

The second part is recognising that vulnerable group outcomes live at the intersection of pastoral care and academic progress, yet these areas can often be artificially separated in school leadership structures. I write honestly as a Deputy Head Academic here. That separation matters, because context shapes learning directly: hearing police sirens in the distance may invisibly claim a pupil’s attention, trauma may affect how pupils process feedback on their work, or a teacher's raised voice may trigger a defensive response that shuts down learning entirely. Remember: your brain’s number one job is to keep you alive. When pastoral and academic leadership treat vulnerable group data as joint evidence rather than separate concerns, the responses are likely to make more sense. The question isn’t whether a child needs emotional support or academic intervention, but how these work together to support the learning most effectively.

The third shift is architectural. If we know that the challenges pupils face may affect neurological function, then robust teaching shouldn't just accommodate this, but be designed around it from the start. This means building in regular checks for listening and understanding rather than assuming everyone's keeping up. It means creating multiple ways for pupils to demonstrate learning, designing lessons where all can participate meaningfully, where thinking is made visible, where no one slips through unnoticed. It means consistent structures that create predictability, clear expectations that reduce uncertainty, and responsive teaching that adapts to pupils’ needs.

Simply put, it’s about recognising that the conditions vulnerable pupils need most are the very conditions that are needed for all. In other words, equity isn’t an add-on; it is good practice made visible.

The leadership test

These shifts ask something specific of school leadership: the willingness to let vulnerable group outcomes interrogate not just individual need, but institutional practice.

The data we generate each term isn't lying to us. But it's only as useful as our willingness to hear what it's really saying about the systems we've created…and our courage to let that signal reshape how we lead.

Its absence is all the stranger when we remember that the very vocabulary of learning is built upon it. Consciousness comes from the Latin con- (with) and scire (to know), literally "knowing with" or shared knowledge. The same root gives us cognition, and reaches back through Indo-European languages to Sanskrit jñāna (knowledge), Greek gnosis, and Germanic kennen. Our educational lexicon is steeped in these roots. Humans have been thinking about thinking for millennia…yet somehow, in education, we avoid naming the thing itself.

Without consciousness no amount of input can become learning. A pupil may hear a lesson, watch a demonstration, even complete a task, but if the experience never crosses the threshold of conscious access, it leaves no trace to be remembered, no foothold for understanding.

Subscribe now

What is consciousness?

Consciousness isn't just philosophical speculation. It's a measurable brain process with distinct dimensions:

• Access consciousness is what matters for learning. Information doesn't just flicker through local brain areas: it "ignites," spreading across networks of prefrontal and parietal regions. Once ignited, it becomes globally available for memory, decision-making, and voluntary action.

• Contents of consciousness are whatever fills awareness at any moment: the specific sights, sounds, thoughts that occupy mental space.

• Levels of consciousness range from alert wakefulness to sleep, anesthesia, or coma: the system's basic on/off states.

Researchers also distinguish phenomenal consciousness: the raw subjective feel of an experience, the redness of red or the pang of pain.

The research shows that only information achieving conscious access gets reliably learned, remembered, and transferred to new contexts.

And these distinctions matter, because they help explain why education depends on consciousness more than it admits.

Why it matters for education

Without naming consciousness, education struggles to examine what it is actually trying to influence. When we say students aren’t “paying attention,” we mean the lesson isn’t reaching conscious access. When we talk about “engagement,” we are concerned with whether content is being consciously processed. When we celebrate “understanding”, we are seeing evidence that information has crossed the threshold of awareness, become available for reasoning, and can now be connected and applied.

Recognising this doesn’t mean every teacher must become a neuroscientist. It’s a design issue. Without the word, education lacks a systematic way of linking classroom practice to research on how learning becomes conscious.

And that research highlights factors that shape conscious access: timing, emotional salience, prior knowledge, competition from other stimuli. These are not abstractions: they are the very conditions teachers already juggle. Yet because the process itself remains unnamed, this knowledge risks staying siloed in neuroscience instead of informing curriculum design, pedagogy, or assessment.

In truth, education already works with consciousness every day. Naming it may simply help us connect practice with the science that explains it.

Why education avoids the word

Its absence has deep roots. Skinner’s radical behaviourism set out to eliminate “mentalistic” concepts altogether: consciousness, thoughts, mental states were declared unscientific. For decades, it was professionally risky to study consciousness openly; the word itself became academically toxic.

By the 1980s, that tide began to turn. Scientists such as Francis Crick and philosophers like Daniel Dennett reclaimed consciousness as legitimate research territory, and psychology moved on. Education, however (always cautious about appearing unscientific) arguably never quite caught up. We learned to talk around the concept, using words like engagement, reflection, or understanding, while avoiding the word itself.

The irony is that consciousness research is now among the most empirical domains of neuroscience, with thousands of studies mapping its neural correlates and measurable processes. And it is not only resurfacing in neuroscience: the very theories that explain how human awareness becomes conscious are now being tested as blueprints for artificial intelligence. When even AI researchers are naming consciousness, can schools afford to keep it unspoken?

Time to name what we mean

Education talks extensively about everything except the mechanism that determines whether teaching works. This isn’t just semantics; overlooking consciousness risks limiting our effectiveness.

Naming consciousness wouldn't solve educational problems overnight. But it may help to connect teaching practice to relevant research, sharpen our understanding of what we're trying to achieve, and help explain why identical lessons sometimes produce learning and sometimes don't.

The word exists. The research exists. The relevance to education is clear. Only the recognition is missing.

Further Reading

Baars, B. J. (2021). On Consciousness: Science & Subjectivity – Updated Works on Global Workspace Theory (N. Geld, Ed.). Nautilus Press

Dehaene, S. (2014). Consciousness and the Brain: Deciphering How the Brain Codes Our Thoughts. Viking.

Mashour, G. A., Roelfsema, P., Changeux, J.-P., & Dehaene, S. (2020). Conscious processing and the global neuronal workspace hypothesis. Neuron, 105(5), 776–798. https://doi.org/10.1016/j.neuron.2020.01.026

What happens when we ask local institutions to solve problems generated at global scale?

Subscribe now

The Institutional Scapegoat Pattern

There's a predictable rhythm to how we handle complex social problems. First, we identify a genuine threat (which is misinformation, in this case). Then comes the policy sleight of hand: hand the problem to schools and leave the "how" to teaching staff who are now expected to become experts in social psychology, media literacy, and digital manipulation tactics.

This isn't unique to misinformation. We see the same pattern with mental health, social inequality, digital citizenship, etc. etc. Schools become the institutional equivalent of a universal solvent: the place where all social problems eventually get dissolved, or at least where we hope they will be.

But this reveals something profound about how we think about institutional responsibility. Schools are attractive as solution-bearers precisely because they're accessible, regulated, and staffed by people with a professional duty to care. They can't simply refuse the problem, deflect to shareholders, or claim neutrality.

They care for children.

The Levels Problem

Misinformation doesn’t spread because children are careless. It spreads because of systems designed to make it spread: algorithms tuned for outrage rather than accuracy, business models that profit from engagement rather than truth, and the wider architecture of social polarisation.

Yet our response is atomistic. The solution is framed not at the level of platform design or regulation, but at the level of individual children: teach them to be more sceptical, more resilient, more careful.

That’s the mismatch: the category error. It’s like trying to solve climate change by improving household recycling while ignoring industrial emissions. The scale of the cure doesn’t match the scale of the cause.

Consider the implications: if misinformation is produced at the speed of algorithms but addressed at the speed of online safety assemblies, if it’s generated by global platforms but countered by local interventions, what does that tell us about our assumptions of how change works?

The Convenience of Displacement

Responsibility often finds its way to schools in a way that feels almost inevitable. It reassures different groups at once: parents see something being done, policymakers can point to action taken, and society has a visible place where The Issue is Being Addressed.

I’d like to think that more often than not, it reflects genuine uncertainty about where responsibility should lie. When problems emerge from the interplay of technology, economics, psychology, and politics….drawing clear lines of accountability is difficult. Schools step into that gap: not because they caused the problem, but because they are one of the few institutions with daily, continuous access to young people and a baseline of public trust.

Yet part of their appeal may also lie in the opposite direction: they are comparatively easy to hold accountable.

But in spite of all this, a deeper puzzle remains. If schools cannot meaningfully address the root causes of misinformation, what are we really asking them to accomplish by a line in policy? And what becomes of institutional integrity when organisations are charged with problems that lie outside their natural scope?

The Adaptation Response

Perhaps the most interesting aspect of how schools actually respond to these impossible asks is that they develop what might be called institutional pragmatism: a kind of learned acceptance of the gap between what's expected and what's possible.

In practice, this means focusing on what they can do: helping pupils recognise rhetorical devices, emotional triggers, and timeless patterns of manipulation. They cannot alter the system that produces misinformation, but they can build resilience in navigating it.

That shift matters. It reframes the task from elimination to inoculation: not eradicating risk, but equipping young people to live within it.

The difficulty is that schools are rarely resourced or supported for this work, yet are held to account as though they alone could close the gap. Without systemic action at the levels where misinformation is produced and amplified, schools risk being praised for adaptability whilst quietly absorbing the cost of a responsibility that was never truly theirs to fully carry.

The Psychology of Institutional Limits

What makes this particularly complex is that the institutional mismatch intersects with psychological realities. Teenagers are fast, intuitive, and deeply social thinkers: exactly the cognitive profile that makes them vulnerable to the ways in which misinformation is weaponised. Funny that. They share content not because they believe it, but to signal identity and belonging. They remember emotional gist rather than precise details. They privilege social proof over logical evidence.

This creates a double bind: the institution being asked to solve the problem (schools) must work within the cognitive constraints that make the problem particularly acute for their population (adolescents). It's like being asked to teach swimming in a pool that someone else keeps draining.

The philosophical question becomes: what do we owe young people when they are caught in problems that arise from the society we’ve made together, but which no single school or individual can put right?

Toward a Philosophy of Bounded Responsibility

Perhaps what's needed is a more honest framework about institutional responsibility: one that acknowledges both the reality of the problems schools face and the limits of what any single institution can accomplish.

This might involve what we could call bounded responsibility: institutions accepting clear accountability for what lies within their sphere of influence while explicitly naming what lies beyond it. Schools can teach pattern recognition, emotional regulation, and critical thinking. They cannot re-design social media algorithms or restructure the tech-bro attention economy.

The Deeper Question

But this raises a final philosophical question: is institutional adaptation to impossible mandates a form of wisdom or complicity? When schools develop sophisticated methods for managing problems they didn't create and can't solve, are they demonstrating admirable resilience or enabling a system that systematically displaces responsibility?

Perhaps the answer is both. The children are real, their vulnerability is immediate, and the problems won't wait for systemic solutions. Schools may be philosophically unsuited to solve misinformation, but they're practically positioned to help young people navigate it.

The misinformation ‘crisis’ may then becomes a lens for examining something much larger: how institutions maintain moral purpose when asked to operate beyond their natural scope, and what we owe each other when individual solutions become the only feasible response to systemic problems.

In the end, perhaps the most honest thing schools can do is teach pupils not just how to spot misinformation, but how to recognise when institutions, including schools themselves, are being asked to solve problems beyond their reach. That kind of institutional literacy might be the most important lesson of all.

Author’s note

This essay began life as something more practical: a set of classroom techniques for tackling misinformation. In writing, I realised I first needed to ask a deeper question about responsibility and scope. The practical strategies (prebunking, teaching manipulation patterns, the “truth sandwich”) will follow in a separate post. This one just needed to lay the ground (/rant!) first.

I run most days. I track it on Strava, set a weekly kilometre goal, and feel a bit rubbish if I don't hit it.

Today was too hot to run, so I tried some strength training instead. Crikey. Not happening. My "fitness" had narrowed to the thing I measure: kilometres. Strava counts kilometres; somewhere along the way I started equating the count with fitness…but they’re not the same thing. And, as it turns out, I now have roughly the upper-body strength of a newborn kitten.

Goodhart’s Law in miniature: when a measure becomes a target, it ceases to be a good measure. My targets started as a nudge towards better health but became the goal themselves. I ended up chasing mileage rather than the fitness it was meant to represent.

And, as ever, there’s a perfect XKCD for this.

That’s the trouble with metrics: they’re useful signals until we start performing for them. Once the number is what matters, we adapt to look good against it. That’s fine if the measure perfectly matches the goal…but most don’t.

And teaching models need similar training and ongoing care to keep them from developing bad habits when no one's watching.

Which, when you drop a new teaching model into INSET, might be why someone in the back will be thinking: "Alright, but what's this really for?" They're right to ask. Teachers and leaders, like everyone else, naturally adapt their behaviour to look good on whatever metrics they're judged by.

And the risk rises when measuring the model creates perverse incentives: when systems unintentionally encourage behaviour that works against their original aim. If teachers are judged on whether steps are visible in a lesson, the safest move is to make them visible, even if that means bending them to fit the context or using them in ways that don't help learning.

This happens because usually it's easier to move the measure than fix the cause, especially when time and energy are tight. It's what gets noticed and what gets recorded, and human nature is to look good when watched.

Perhaps that's why I've been exploring Steplab’s Simple Model of Teaching so closely these past weeks…To see not just what it offers, but what might happen when a good teaching model starts to wander: how far a solid idea could drift from the cognitive science that shaped it, and whether we can train it to stay true to its purpose.

What this series has explored

Over the past three posts, we've explored what the Simple Model of Teaching can't easily display: the cognitive and neural processes that drive learning. Short-term memory serves as the fragile hinge between perception and manipulation, shaped by language, safety, and prior knowledge, amongst others. Working memory provides the cramped workspace where thinking happens, easily overloaded or disrupted. Long-term memory involves the slow strengthening or fading of learning through sleep, retrieval, spacing, and curriculum design.

These processes are complex, conditional, and often invisible in the moment…yet they're at the heart of how learning happens. They're also what make teaching models so valuable…and so vulnerable to mishandling.

Subscribe now

When frameworks meet systems

Steplab's Simple Model of Teaching is built from these cognitive insights. But once it enters a school system, it will inevitably be translated into what can be observed and tracked. That translation often happens with the best intentions: to support consistency, make improvement visible, and create shared language.

The problem emerges when the model gets expressed in metric-friendly form. The underlying cognitive science may start to flatten unless it's kept alive. Areas still actively debated in research get simplified for practical use, and then perhaps fixed into metrics that can harden into practice long after the evidence has moved on.

This is where good models can go bad. Without careful training, they learn to optimise for measurement rather than learning, like a dragon that's figured out how to hoard data points instead of breathing pedagogical fire.

So how do we train our teaching models to resist these pressures? First, we need to recognise the warning signs when well-intentioned frameworks start developing bad habits.

Here are four symptoms of an untrained model you may have encountered:

• Causal Goodhart – Mistaking the indicator for the cause. This happens when an indicator (something that tends to appear when learning is happening) is treated as if it directly produces learning. The logic is: if we make the indicator happen, the outcome will follow. The underlying aim may be harmed as a result.

  Example: Research shows effective teachers often use cold calling. A school mandates that every observed lesson must include it. Teachers start cold calling mechanically, without the scaffolding or classroom culture that make it work. The behaviour is present, but the underlying cause of its success is missing.

• Regressive Goodhart – You select people or results because they look extreme on a noisy measure; on re-test they drift back toward typical levels, creating the illusion of improvement or decline.

  Example: A January wellbeing survey flags a small group with unusually low wellbeing. The school runs a four-week booster; in March their scores improve. Leaders credit the programme, but some of the bounce was always likely…those pupils were selected at a low point on a noisy measure and would likely have lifted on re-test even without the intervention.

• Lethal mutation – The simplified, measurable version becomes what gets remembered and passed on. Original cognitive principles may remain true, but they're no longer visible in daily practice.

  Example: Dual coding research shows that combining verbal and visual information can aid learning. This evolves into “every lesson must include visuals.” Teachers start adding decorative images or busy infographics that increase cognitive load rather than reduce it.

• Ossification – Once written into policy or accountability frameworks, the ‘metricised’ form locks in place. Even when science or understanding evolves, practice stays frozen because the system rewards the old shape.

  Example: A trust implements "trauma-informed" behaviour policies in 2020 based on early research about adverse childhood experiences. New evidence emerges about the complexity of trauma responses and the risk of over-pathologising usual stress responses. But the policies are now written into safeguarding frameworks and teacher training modules. Even as understanding of trauma and resilience becomes more nuanced, schools stick with simplified "trauma-informed" checklists because updating would mean overhauling established systems.

None of this happens through malice. These shifts emerge through the natural pull toward what's easiest to notice, record, and share. Left unchecked, they leave us with a framework that looks identical on paper…but has lost its teeth.

If we want them to stay sharp, we have to treat their use as an ongoing discipline, not a one-off launch.

How to Train Your Teaching Model

Like any training program, this requires accepting that setbacks are inevitable. Models, like pupils, will occasionally misbehave despite our best efforts. The challenge is that you’re not training them in a vacuum…you’re training them inside a system that runs on measures.

Schools face an impossible equation: they need metrics to function at scale, but learning happens through cognitive processes that arguably resist tidy measurement. We can't resolve this tension by wishing it away or pretending one side doesn't matter.

Instead, we need to accept that drift is inevitable. The pressure to turn frameworks into checklists isn't bureaucratic incompetence: it’s how large organisations survive. Schools must allocate budgets, justify decisions to governors, and demonstrate progress to parents and inspectors. A nuanced understanding of working memory doesn’t always help when you’re expected to evidence progression in a five-minute walk-through.

So, frameworks will always get simplified. The question isn't whether this will happen, but whether anything meaningful survives the translation.

Design for graceful degradation

Some models are naturally more trainable than others. The Simple Model may need extra care because its steps look simple while depending on complex, invisible processes. Better might be frameworks that do the following:

• Build uncertainty into the structure itself. Instead of "ensure retrieval practice happens", try "experiment with these types of retrieval methods and track what works over time". Train your model to expect variability rather than demand consistency. The latter is harder to tick off a checklist because it explicitly acknowledges that the work is ongoing and contextual.

• Make the complexity visible in the simplification. Rather than hiding the messiness of working memory research behind clean steps, embed the uncertainty: "Support working memory (but remember: we're still learning how this works, and it varies between pupils and contexts)". Good training doesn't hide the difficulty…it prepares you for it.

• Include natural feedback loops. Build in requirements to check back on the things that matter most but show up latest. Train your model to look for delayed results, not just immediate compliance. If the model includes a step about building long-term retention, make it impossible to "complete" without evidence from weeks or months later, like checking your actual fitness rather than just counting kilometres.

Another safeguard is to design clear boundaries around a model’s purpose. Without them, it can be pulled into roles it was never built for.

Separate frameworks from accountability

I recognise that Steplab does guard against this, but perhaps the real mistake schools or trusts might make is assuming the same tool can serve two masters: learning and accountability. The Simple Model of Teaching might work well as a thinking tool for teachers and coaches whilst being entirely unsuitable for performance management or inspection frameworks.

That distinction should be made explicit and built into the model's training from day one. Don't let your teaching framework learn that performance management is its primary purpose.

And it suggests a different approach: develop parallel systems. Use rich, complex frameworks for professional development and reflective practice. Use more robust (and maybe cruder?) metrics for accountability.

And, as always, the risk is that the accountability metrics will still distort practice. But at least the professional learning frameworks can remain intact, creating space for deeper understanding even within measured systems.

Even then, we still face the harder challenge: some of the most important outcomes in learning can’t be captured by simple metrics at all. Protecting the framework is one thing; finding ways to notice and track those outcomes without flattening them is another.

Measure the unmeasurable differently

Some learning outcomes may resist direct measurement but respond well to patient observation:

• Longitudinal patterns – “How has this pupil’s ability to plan and complete complex tasks changed over the term?”

• Negative indicators – Track signs of overload (confusion spikes, recurring error types, disengagement) rather than only logging positive moves.

• System-level effects – Track whether pupils are retaining learning across curriculum boundaries, whether they're making connections between topics, whether they're developing increasingly sophisticated thinking.

These approaches train your measurement systems to look for depth rather than just surface performance. They're harder to game because they require sustained, authentic learning rather than temporary compliance.

Keep the tension alive

Maybe the measurement paradox isn't a problem to solve but a training routine to maintain. The pull between understanding and accountability might be like resistance training: uncomfortable, but strengthening.

Teachers and leaders should feel this tension as productive feedback, not failure. Models that never feel any resistance probably aren't being challenged enough to grow. A teaching framework that sits too comfortably in your context might have stopped learning.

In this vein, some friction is not just inevitable but useful. Teachers should feel the creative tension between what the model suggests and what their context demands. Leaders should feel the productive discomfort of knowing their metrics don't capture everything that matters. That discomfort signals the system is still alive to the complexity of learning.

The goal isn't to eliminate this tension but to make it work for you. Use the friction between measurement and meaning as a source of ongoing inquiry: not a problem to solve once and for all, but a dynamic that keeps both sides honest.

Training a teaching model isn't a one-time implementation; it's building an ongoing relationship. Like any worthwhile training, it requires patience, consistency, and the wisdom to know when to celebrate progress and when to course-correct. Some days your model will perform beautifully. Other days it will chase the wrong metrics or forget its core purpose. Both are part of the process.

Steplab’s Simple Model of Teaching works because it emerges from solid cognitive science. Keep it working by treating it as a companion in learning, not a rigid master. Stay curious about what it's teaching you, remain sceptical when it gets too comfortable, and remember that the best models, like the best teachers, never stop learning themselves.

References

Goodhart, C. A. E. (1975). Problems of monetary management: The U.K. experience. In A. S. Courakis (Ed.), Monetary theory and practice (pp. 91–121). London: Macmillan.

Munroe, R. (2024, February 26). Goodhart’s law [Comic]. xkcd. https://xkcd.com/2899

Strathern, M. (1997). ‘Improving ratings’: Audit in the British University system. European Review, 5(3), 305–321.

Wheeler, D. J. (1993). Understanding variation: The key to managing chaos. SPC Press.

In previous posts, I’ve explored short-term and working memory and how teaching skilfully means knowing about and working within these cognitive limits.

Steplab’s Simple Model of Teaching (above) sketches the key things teachers should think about: selecting curriculum, securing attention, presenting ideas clearly, driving thinking, giving feedback, and consolidating knowledge.

And the more I’ve worked with this model and explored it, the more I’ve realised something essential is missing.

Time.

Effective teaching, and lasting learning, needs time. Time to think, to forget, to revisit, to connect. Not just more of it, but a better recognition of it.

Time can’t be rushed. But it can be planned for.

Subscribe now

Seconds and minutes: A fragile beginning

At the end of a lesson on tectonic plates, a pupil gave a near-perfect explanation of subduction. He used the correct terminology and, to the class’s delight, acted it out. It felt like one of those moments where everything clicks. Brilliant! We can move on.

But when I asked him again a week later, he faltered.

That isn’t unusual. And it’s likely not a sign of laziness, distraction, or poor teaching. It’s just how memory works.

In the first minutes after learning something new, the brain responds with a rapid burst of activity. This activity creates a synaptic tag, which is a short-lived signal that flags those neural connections as worth keeping. But unless something follows to reinforce that signal (such as a revisit, a retrieval task, a new connection) the tag fades, usually within a few hours.

I’m sure you can see the problem here.

So by the time a pupil gives a confident answer at the end of a lesson, we may be seeing short-term performance rather than long-term retention. That moment of clarity hasn’t yet had the chance to settle into memory.

And this has implications. It reminds us that:

• Retrieval within the lesson matters, even a quick mid-lesson check-in can strengthen fragile traces.

• Distractions or sudden changes can interrupt consolidation before it begins.

• Most importantly, we shouldn’t assess too soon. Without a revisit, what seems understood today may slip quietly away tomorrow.

Hours and days: The night shift

Someone recently floated the idea at our school’s wellbeing forum that we should allow naps during the day. People laughed, but perhaps they shouldn’t have. The brain, it turns out, does some of its best work while we sleep. Good news for colleagues on their summer holidays.

In Early Years settings, rest is protected as we recognise its developmental power. But as children grow older, rest is often replaced by rigour. We tend to fill every available hour with new input and new tasks, expecting focus from children through six to eight hours of new input without pause.

This is even as the biological evidence keeps mounting.

Memory doesn’t just form during a lesson. It forms after, especially during sleep.

Sleep is the brain’s second shift, coordinating replay and integration. During deep sleep, the hippocampus rapidly reactivates traces of the day’s learning in patterns known as sharp-wave ripples. These fleeting bursts help co-ordinate activity across the cortex, loosely reactivating learning traces and gradually binding them into longer-term memory networks.

But here's the catch: sleep replay isn’t always clean or complete. As Findlay et al. (2021) explain, it’s often noisy and fragmented. Replay doesn’t just consolidate: it also transforms, linking memories across time and space. Without it, learning remains fragile. One night of sleep deprivation can leave new knowledge harder to retrieve, more easily forgotten, and less likely to stick.

What does this mean practically?

• Support consolidation, not overload. Homework doesn’t need to introduce new input. Short retrieval tasks, reflections, or rehearsals can help strengthen traces laid down earlier, without competing for attention or sleep.

• Time assessments with memory and recovery in mind. Where possible, avoid clustering big tests at the end of a day or just before weekends which may be times when pupils are most likely to cram and least likely to sleep well.

• Respect recovery windows. The brain needs quiet space to integrate.

• Talk about sleep as part of learning. Make rest normal. Not indulgent.

Days and weeks: The quiet beauty of forgetting

At my school, we’ve been living through a building project, temporarily working out of a block we call Ichthys. The name sounds grand, but in truth it’s a tired 1960s CLASP science block we had to evict the pigeons from. Neither the building nor its landscaping was built for five hundred energetic children (let alone the staff…but let’s not go there).

As a result, the grassy areas around the building have become crisscrossed by narrow trails, etched by the feet of pupils and staff choosing routes that make sense to them.

These paths, like the one shown below, are known as desire paths.

And memory, in some ways, works the same way.

Each time we teach something new, we trace a tentative path through our pupils’ minds. At first, it’s easily overgrown. Leave it untouched too long, and it vanishes. Revisit it too often, and pupils never need to find it themselves. Fail to connect it to other paths, and they’ll struggle to explore related ideas.

But if we return to an idea just as it’s beginning to fade, pupils have to work a little harder to recall it…and that effort helps strengthen the memory, deepen the path.

And this is the quiet paradox of forgetting: to strengthen memory, we have to let it fade a little.

Retrieving a partially faded memory signals to the brain: This matters. Keep it.

What does this mean practically?

• Plan retrieval sessions across time. Don’t just revisit ideas the next day - return to them after a week, a month, a term.

• Let pupils work for it. Effort signals strengthening, not loss (Bjork & Bjork, 2011).

• Link old paths to new ones so networks grow richer.

Below is a simple example of how retrieval can be spaced across a unit on Coasts:

Weeks and years: When curriculum lifts the weight

Over weeks and years, repeated activation reshapes how and where a memory is stored.

Early on, memories rely on the hippocampus. Over time, reactivation spreads the load across the prefrontal cortex, parietal lobes, sensory cortices, and amygdala. The hippocampus often remains involved, but the network becomes broader and more flexible.

This slow integration is why curriculum matters. We’re not just reviewing content. We’re shaping the architecture of memory.

Too often, topics are taught once, assessed, and left to fade. Weeks or months later, pupils reach for them and find fragments. Even a brief return - a quiz, a recap, a cross-topic link - can re-awaken the network and preserve the cross-links that make ideas usable.

One concept, revisited across a year, is more likely to endure than one crammed into a single block. But these moments don’t happen by chance: they depend on planning across subjects as well as within them.

A persuasive structure in English might resurface in a PSHE speech. A Year 7 musical scale might return in Year 9 composition, or echo in mathematical pattern recognition. These echoes deepen and connect memory.

Emotion strengthens it too. When the amygdala is engaged through curiosity, novelty, or relevance, it boosts hippocampal encoding. Well-timed stories, questions, or surprises can make ideas stick in ways a worksheet can’t.

But it has to be real. The aim is for pupils to remember the learning, not just the stunt. A clever hook that overshadows the concept risks leaving them with the punchline, but not the point.

Implications?

• Sequence curriculum so key concepts resurface across months and years, not just within a single unit.

• Plan for cross-subject echoes - co-ordinate so ideas taught in one area reappear in another.

• Pair retrieval with relevance - link revisits to curiosity, novelty, or authentic context to strengthen encoding.

• Avoid “memorable” activities that overshadow the learning: the goal is for pupils to recall the concept, not just the stunt.

Conclusion: Why curriculum completes the model

The Simple Model of Teaching demonstrates effective practice inside a lesson: it gives teachers a teaching playbook for steering short-term and working memory.

Yet durable learning relies on what follows: minutes later, hours later, days later, and even years later.

A well-sequenced curriculum:

• Allows just enough forgetting to make retrieval powerful.

• Spaces returns to deepen understanding across contexts.

• Builds emotional salience and variation to embed ideas.

That’s why, alongside our models of teaching, we also need clear models, and a shared language, for curriculum development. Deciding what to revisit, when, and why is intellectually demanding, creative work carried out quietly by Heads of Department and curriculum leads.

Curriculum sequencing lifts the floor. Lesson-level coaching raises the ceiling. Together, they give every pupil - not just the most prepared or privileged - the best chance to learn well, and remember deeply.

References

Bjork, E. L., & Bjork, R. A. (2011). Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. In M. A. Gernsbacher, R. W. Pew, L. M. Hough, & J. R. Pomerantz (Eds.), Psychology and the real world: Essays illustrating fundamental contributions to society (pp. 56–64). Worth Publishers.

Findlay, G., Tononi, G., & Cirelli, C. (2021). The evolving view of replay and its functions in wake and sleep. Sleep advances : a journal of the Sleep Research Society, 1(1), zpab002. https://doi.org/10.1093/sleepadvances/zpab002

Mccrea, P. (2024, January 17). A simple model of teaching [Tweet]. Twitter/X. https://twitter.com/PepsMccrea/status/1747685287633086862 Image adapted from original by Oli Cav.

Moncada, D., Ballarini, F., & Viola, H. (2015). Behavioral Tagging: A Translation of the Synaptic Tagging and Capture Hypothesis. Neural plasticity, 2015, 650780. https://doi.org/10.1155/2015/650780

Nicoll R. A. (2017). A Brief History of Long-Term Potentiation. Neuron, 93(2), 281–290. https://doi.org/10.1016/j.neuron.2016.12.015

In my previous post, I explored short-term memory: the cognitive process that holds information briefly so that thinking can begin, and the one not labelled on Steplab’s Simple Model of Teaching (Figure 1).

Subscribe now

But, this post will focus on one process that is labelled: working memory.

The fact that it’s labelled should make things simpler. But it doesn’t.

Despite being central to the diagram (nearly every arrow leads to or from it) there’s not consensus on how working memory…well…works.

This was a difficult post to write. There’s a lot of literature on working memory, and a number of competing models. So in this post, my aim is simply to sketch what it’s generally thought to involve - drawing on cognitive science and a bit of neuroscience - so we can better understand, and support, the pupils we teach.

In this field, scratch the surface and one model tends to lead to another.

It’s models all the way down.

What exactly is working memory?

Imagine this:

You’re taking your class on a trip to the zoo.

You’ve just stepped out of the butterfly house.

It’s warm, humid, and you’re recovering.

The pupils are lining up for the next enclosure.

You run a quick head‑count: twenty‑seven.

But you know there should be twenty‑eight.

You scan the group and picture where they stood a minute ago. That mental snapshot relies on your visuospatial sketchpad: the part of working memory that holds visual and spatial information.

As you count again, you whisper their names to yourself (Ella, Leo, Judah …) using your phonological loop, which maintains verbal information.

Meanwhile, your central executive runs the show: staying focused, switching strategies, holding back the urge to shout, “Who’s missing?”

It’s Ava, now sprinting up with a butterfly perched triumphantly on her finger.

You breathe out, relieved.

Baddeley and Hitch’s model of working memory

The headcount in the butterfly house didn’t rely on a single ‘box’ in the brain.

It drew on a set of temporary systems, each doing something slightly different: visual, verbal, and attentional.

This is the core idea behind one of the most influential models of working memory proposed by Baddeley and Hitch in the 1970s. Earlier models proposed that information would make its way into long-term memory if it just sort of brewed in short-term memory long enough, as if time alone would do the work…like a good cup of tea (which I hope you’re drinking as you read this).

But evidence pointed to the contrary.

And so Baddeley and Hitch (1974) argued that working memory isn’t a single store, but a system of components….a process, as displayed in Figure 2:

• Phonological loop holds verbal or auditory information for a few seconds and refreshes it through rehearsal

• Visuospatial sketchpad stores what’s seen and where it is

• Central executive is not a store, but a controller: it shifts attention, switches strategies, and manages the flow of information across the system. It has, however, been criticised for being a bit homunculus…y

This framework doesn’t explain everything, especially not how attention actually works, or how meaning is integrated, but it’s a useful starting point. It helps us name & consider the different types of load we’re asking pupils to carry.

Sadly, it’s not that simple

Think back to Ava’s butterfly.

You likely pictured a whole butterfly and not separate patches of colour, shape, and motion. You may have had a visual image, an emotional response, even a flicker of memory. Unless, of course, you have aphantasia…in which case, your mental recall may have taken another form entirely.

The point is: how does the brain bind all these features…visual, emotional, spatial…into a single, coherent image?

This is known as the binding problem, and it’s still unresolved. For years, it challenged the original three-part model of working memory. Neither the phonological loop nor the visuospatial sketchpad could explain how we combine different types of information, or why we recall meaningful sentences more easily than lists of random words.

To address this, Baddeley (2000) introduced a fourth component to the model: the episodic buffer. It’s a temporary store that integrates information from different sources into a unified whole.

Initially thought to be controlled by the central executive, later research suggests that binding may emerge from more domain-specific processes, such as visual attention or language networks, rather than from a single overseer.

Where does that leave us? With a complex system, still under study. But with a clearer sense that working memory draws on multiple systems, each with its own limits, and that integration isn’t effortless.

And that has consequences.

Can pupils multitask?

You might think you’re good at it…and maybe you are. But statistically, frequent multitaskers are more distractible, not less. They tend to spread their attention more broadly, but with less control over where it lands.

Multitasking may feel efficient. But it comes at a cost, even for adults. For children whose working memory systems are still developing, the cost is often higher and recovery slower.

We often assume pupils can juggle multiple streams: write while listening, read while someone talks, take in new instructions mid-task. Why else would we keep talking as we set them off to work?

But, no, they can’t.

As we’ve seen, the most influential model of working memory involves separate systems including the phonological loop, the visuospatial sketchpad, the central executive. These can function in parallel, but when two tasks rely on the same system, they tend to interfere.

It’s why reading text from a PowerPoint whilst trying to listen to something different being said is so difficult. You focus on one or t’other. Both use the phonological loop. The same system is trying to process two streams at once.

Studies on dual-task performance show that the more similar two tasks are (in modality or response type) the more likely they are to compete. It’s why we can drive and sing (or try to…), but not drive and text. A pupil holding a sentence in mind while writing it down, and then being spoken to mid-task? Something gets lost, not through lack of effort, but because the system is overloaded.

In short, shared resource = more interference.

Separate resource = more manageable…but still not risk-free if one task becomes demanding (why people may stop singing when trying to park).

In practice, multitasking often means task-switching. And switching interrupts encoding, fragments attention, and displaces what was already held in mind.

This matters even more for pupils with EAL, pastoral needs, ADHD or dyslexia, who may already be rehearsing more, retrieving more slowly, or compensating for reduced working memory capacity.

For these pupils, an interruption isn’t a gentle redirection. It’s an eviction.

So the rule of thumb holds:

Break tasks into steps. Avoid talking over writing, reading, or movement.

Not just for clarity, but to preserve the scarce workspace they’re already using to think.

And sometimes, that workspace is under pressure before we even begin.

Working Memory Hijackers

Working memory isn’t just limited - it’s vulnerable.

Imagine trying to work from a tiny desk.

Now add clutter.

Below are five common intrusions, drawn from recent research, and what they might look like in the classroom.

What does neuroscience add?

It confirms what many teachers sense: working memory isn’t just limited…it’s fragile.

Neuroscience shows that working memory depends on a network of brain regions working together. The prefrontal cortex helps us hold and manipulate information; the parietal cortex helps direct attention and spatial awareness. These areas don’t work alone…they’re supported by subcortical structures such as the basal ganglia, thalamus, and cerebellum. When any part of that system is overloaded, whether that’s by noise, stress, or ‘multitasking’, performance drops.

Importantly for children, this network is still developing. The neural connections involved in working memory are slow to mature. The pathways that support attention and control are among the last to fully myelinate. This means children's capacity isn’t just smaller than adults'.

It's more easily disrupted.

Neuroscience also helps explain why information sometimes “comes back” with a cue. Some memories are kept active by ongoing brain activity, but others are held in a quieter, background state…still there, but harder to access without support. A calm prompt can reactivate what seemed lost; a sudden interruption can wipe it out completely.

So while cognitive models describe what working memory does, neuroscience helps us understand why it’s so sensitive and why some pupils may need more time, space, and support just to keep a thought in mind.

Where does this leave us?

This was a difficult post to write.

Working memory sits at the centre of the Simple Model of Teaching (Figure 1). Almost every arrow points to or from it, and yet it really & truly resists simple description. It’s not one thing, or in one place. It isn’t neatly bounded, nor fully agreed upon. There are competing and conflicting models, overlapping systems, and findings that clarify and complicate in equal measure. A great deal has been left out.

But that’s been kind of deliberate. The goal here wasn’t to cover everything. It was to look beyond the label and sketch what working memory broadly involves and to, hopefully, make it easier to teach with it in mind.

Because if we treat working memory as a single box in the model, we risk missing what’s inside it: the juggling, the rehearsal, the effort, the load. The pupils.

The model moves on from here: to practice, to retrieval, to long-term memory, to forgetting. It’s important to remember that working memory isn’t a passing step though. It’s the workspace of learning, and it needs protecting.

References

Baddeley, A.D., & Hitch, G. (1974). Working memory. In G.H. Bower (Ed.), The psychology of learning and motivation: Advances in research and theory (Vol. 8, pp. 47–89). New York: Academic Press.

Baddeley A. (2000). The episodic buffer: a new component of working memory?. Trends in cognitive sciences, 4(11), 417–423. https://doi.org/10.1016/s1364-6613(00)01538-2

Chai, W. J., Abd Hamid, A. I., & Abdullah, J. M. (2018). Working memory from the psychological and neurosciences perspectives: A review. Frontiers in Psychology, 9, Article 401. https://doi.org/10.3389/fpsyg.2018.00401

Geißler, C. F., Friehs, M. A., Frings, C., & Domes, G. (2023). Time-dependent effects of acute stress on working memory performance: A systematic review and hypothesis. Psychoneuroendocrinology, 148, 105998. https://doi.org/10.1016/j.psyneuen.2022.105998

Gheller, F., Spicciarelli, G., Scimemi, P., & Arfé, B. (2024). The Effects of Noise on Children’s Cognitive Performance: A Systematic Review. Environment and Behavior, 55(8-10), 698-734. https://doi.org/10.1177/00139165241245823 (Original work published 2023)

Jiahui Li, Yixuan Cao, Simei Ou, Tianxiang Jiang, Ling Wang, Ning Ma, The effect of total sleep deprivation on working memory: evidence from diffusion model, Sleep, Volume 47, Issue 2, February 2024, zsae006, https://doi.org/10.1093/sleep/zsae006

Ma, W. J., Husain, M., & Bays, P. M. (2014). Changing concepts of working memory. Nature neuroscience, 17(3), 347–356. https://doi.org/10.1038/nn.3655

Mccrea, P. (2024, January 17). A simple model of teaching [Tweet]. Twitter/X. https://twitter.com/PepsMccrea/status/1747685287633086862 Image adapted from original by Oli Cav.

Pickering, H. E., Parsons, C., & Crewther, S. G. (2022). The effect of anxiety on working memory and language abilities in elementary schoolchildren with and without additional health and developmental needs. Frontiers in Psychology, 13, Article 1061212. https://doi.org/10.3389/fpsyg.2022.1061212

von Hippel, C., Kühner, C., Coundouris, S. P., Lim, A., Henry, J. D., & Zacher, H. (2024). Stereotype Threat at Work: A Meta-Analysis. Personality and Social Psychology Bulletin, 0(0). https://doi.org/10.1177/01461672241297884

As we prepare to launch Steplab across our school this autumn, we’ve been spending time with its Simple Model of Teaching (Figure 1): a clear, structured framework that supports planning, coaching and professional development. It’s designed to help teachers focus on what matters most and it’s a helpful tool. But like any model, it works best when we bring our own knowledge to it…not just of learning, but of learners.

This series is an attempt to do just that. And writing for me is a way of thinking aloud and an opportunity to reflect on how best to implement this work before we roll it out. I am grateful that the summer holidays, in between time with friends, family…and Expedition 33 (!), offers space for that kind of reflection.

So, this series aims to be a companion to the Simple Model of Teaching. A way of thinking alongside it and of filling it out.

We begin with short-term memory: a step that doesn’t appear in the diagram, or the one it’s adapted from, but one that underpins the rest.

It’s where deliberate learning begins.

What is short-term memory?

Short-term memory (STM) refers to the brief, temporary holding of a small amount of information, typically for just a few seconds. It’s limited in both capacity and duration and unless the information is rehearsed, it tends to fade quickly.

Some models treat STM as a separate system; others see it as overlapping with working memory, which I’ll explore in my next post. Either way, the distinction matters. Unlike working memory, which involves actively manipulating information, short-term memory simply holds it in mind. It doesn’t solve the problem; it just keeps the pieces available.

STM is what allows a child to:

• remember the first part of your sentence while you’re still speaking the last

• copy a diagram from the board

• hold the digits of an equation just long enough to write them down

It’s often overlooked, as it has been in the model we’re discussing, but if information isn’t held, it can’t be worked with.

Everyone stores the world differently

When we look out at the world, we don’t see the world.

We perceive a model of the world.

And we all have our own models shaped by experience, expectation, and what we notice.

Helpful when you’re trying to get 30 different kiddos with 30 different models of the world to understand the same 1 model of tectonic plates.

Because short-term memory doesn’t store objective reality. It stores what was attended to, understood, or rehearsed. In that sense, it’s as much about perception as it is about memory.

And perception isn’t neutral.

Two children hear the same instruction. One acts. One hesitates. It’s not always about motivation or comprehension. It might be about what was stored, and how.

And that varies.

What if the classroom feels unsafe?

There’s an image I keep coming back to: a comparison of eye-tracking scans between a non-artist and an artist viewing the same scene. The difference is striking. The artist’s gaze moves widely, taking in the periphery, the shape of negative space. The non-artist’s gaze clusters tightly around the central subject. See Figure 2, below.

Now imagine that same idea in the classroom.

Some children, particularly those who’ve experienced trauma, may not scan the room to take in information. Instead, without even realising, they scan for threat.

Eye-tracking research shows that people with trauma and anxiety are drawn rapidly and automatically toward threat and once their attention lands, it struggles to shift. See Figure 3 below. The system isn’t faulty. It’s protective.

And if attention is pulled toward risk, then short-term memory follows. In cognitive theory, STM typically only stores what has been perceived, and perception is where attention goes. It’s not defiance, or carelessness, or poor listening, or poor memory. It’s the brain doing exactly what it learned to do: watch for danger.

The system’s not failing.

It’s protecting.

STM isn’t measured in a vacuum

In research and assessment contexts, short-term memory is often measured using tasks such as digit span, where a pupil is asked to repeat a string of numbers in order. It’s a simple way to estimate how much verbal information can be held briefly in mind.

It’s relevant: the WISC-V (Wechsler Intelligence Scale for Children) uses digit span to help measure STM.

But like all assessments, digit span isn’t neutral. It reflects not just memory capacity, but also language familiarity, cultural experience, and strategies shaped by context.

This matters, especially when we’re working with pupils from diverse linguistic backgrounds.

It’s well established that digit span varies across languages, and that this variation is partly explained by the spoken length of number words. In some languages, such as Mandarin, number words are shorter and quicker to articulate, than, Welsh, for example, which makes them easier to rehearse in the mind before they fade. The faster we can repeat something to ourselves, the more of it we can keep hold of.

In cognitive terms, that silent repetition is known as subvocal rehearsal, or more informally, the “inner voice”. It’s part of the phonological loop, and it’s one of the main strategies we use to maintain information in verbal short-term memory. It’s probably not very effective as a strategy, but that’s beyond the scope of this post.

But it’s not just about speed.

Research suggests that some language groups also draw on different memory strategies (Baddeley, Xu, Ho, & Hitch, 2023). A striking feature of speakers of Mandarin and Cantonese is the fact that their immediate verbal memory span tends to be substantially greater than is found for other languages. For example, speakers of Mandarin and Cantonese continue to show a phonological similarity effect (the tendency to confuse similar-sounding words) even when silent rehearsal is blocked. While this effect typically disappears under articulatory suppression, its persistence here suggests that they may be drawing on more than just an “inner voice.” They appear to use an “inner ear” as well and maintaining sounds through auditory imagery as well as articulation.

This dual coding (drawing on both articulation and auditory imagery) may reflect the demands of learning a complex writing system. But it also reminds us that verbal memory strategies are not universal. They’re shaped by language, culture, and the ways we’ve learned to process sound and meaning.

For pupils learning English as an Additional Language (EAL), especially those still developing fluency, there is often an additional processing demand at the point of input. Before verbal information can be stored, it first has to be decoded, and when the language of instruction is unfamiliar or still being mapped, that decoding process draws on valuable cognitive resources.

It means that, in practice, working memory is being engaged earlier - just to access the input - before short-term memory can even begin to hold it.

It does mean that tasks which assume instant access to language, whether in assessment or classroom dialogue, can underestimate how much effort is already being expended just to hear, segment, and interpret what’s being said.

Supporting children at this stage means recognising that rehearsal and retention are not always immediate. Clear, paced input and space for repetition aren't just helpful: they make memory (and therefore learning) possible.

Where short-term memory fits

The Simple Model of Teaching doesn’t show short-term memory.

And yet, in practice, it’s everywhere. It sits inside every arrow, every step, every well-chosen question. Before a pupil can work with an idea, they must first hold it, even if just briefly. And that brief holding depends on more than just teaching technique, as this post has begun to explore. There’s even more to it…but that’s for another time.

If we were to make space for short-term memory in the model, I wouldn’t wedge it in as a separate box. I’d thread it through, perhaps between what the model labels as “securing attention” and “optimising communication”.

It’s like the hinge between perception and manipulation.

And, referring to the diagram, when we reduce that moment to a question like “Has the teacher got students’ attention?” we risk flattening it. That phrase makes attention sound simple, as if it is something the teacher either captures or doesn’t. It implies control. Compliance. That everyone is starting from the same place.

But attention is not a switch. It’s variable, fragile, and contingent on emotion, language, sensory access, fatigue, and context. And it certainly doesn’t always move through the eyes, despite what diagrams (or practices such as SLANT) might suggest.

A better question might be:

Has attention been invited, supported, and sustained?

Or:

Are conditions in place for attention to be possible?

These kinds of questions shift the gaze from control to co-regulation.

The same goes for curriculum. “Has the teacher selected the right ideas to teach?” risks sounding hierarchical as if the job is simply to choose, rather than to connect.

We might ask instead:

Do the ideas connect to prior knowledge, lived experience, or future learning?

These are more than semantic tweaks. They shape how we interpret what we see. And they remind us that models are starting points, not prescriptions.

Short-term memory isn’t always something we can observe directly.

But when we think with it in mind, we tend to notice more.

And that often changes what we do next.

If you'd like to keep reading...

The next post will look at working memory. You can subscribe by clicking the button below to receive it directly when it’s published.

Subscribe now

Thank you for reading.

References

Armstrong, T., & Olatunji, B. O. (2012). Eye tracking of attention in the affective disorders: A meta-analytic review and synthesis. Clinical Psychology Review, 32(8), 704–723. https://doi.org/10.1016/j.cpr.2012.09.004

Baddeley, A. D., Xu, Z., Ho, S. T., & Hitch, G. J. (2023). On verbal memory span in Chinese speakers: Evidence for employment of an articulation-resistant phonological component. Journal of Memory and Language, 129, 104389. https://doi.org/10.1016/j.jml.2022.104389

Mccrea, P. (2024, January 17). A simple model of teaching [Tweet]. Twitter/X. https://twitter.com/PepsMccrea/status/1747685287633086862 Image adapted from original by Oli Cav.

Vogt, S., & Magnussen, S. (2007). Expertise in pictorial perception: eye-movement patterns and visual memory in artists and laymen. Perception, 36(1), 91–100. https://doi.org/10.1068/p5262

There’s a diagram doing the rounds, see Figure 1 above.

Six neat steps, boxed and arrowed, moving from curriculum choices through to consolidation. It’s Steplab’s (2024) Simple Model of Teaching and it’s being used increasingly in schools to shape planning, delivery, and feedback. At my school, we’re starting to use this model more explicitly in our professional development & I think it’s a really helpful tool. Because in many ways, it works.

It captures effective teaching: focus attention, reduce unnecessary load, revisit what matters, and build towards long-term retention. It draws on cognitive science without being heavy-handed, and it gives a shared language for talking about what happens in the classroom.

Boom. Teaching, explained.

Or is it?

Subscribe now

Because teaching isn’t simple.

And neither is memory.

But simple models aren’t meant to capture everything. They’re meant to give us something to stand on. And once you’re standing, you can start to look around a little.

Teaching and Memory

At the centre of the model is working memory. It’s shown, rightly, as a bottleneck: the narrow space through which attention and understanding must pass before anything can be stored. The model encourages us to design instruction with this in mind to avoid overload, to teach deliberately, to help pupils retrieve what they’ve learned.

The diagram is only a surface though. There are processes it doesn’t show & systems that sit just beneath the arrows. For example:

• When a child forgets part of your instruction after ten seconds, that’s not working memory. That’s short-term memory, and it isn’t on the diagram.

• When pupils appear to follow your modelling but can't repeat the steps, that might not be misunderstanding - it might be the limits of sequence storage.

• When they seem to grasp something and forget it days later, that isn’t a failure of attention - it’s memory doing what it’s designed to do. Forgetting isn’t a bug in the system. It is the system.

None of this undermines the model…though it does remind us that models simplify. That’s their job. And that’s also the risk.

What This Series Is (and Isn’t)

In the next few posts, I’ll be looking at:

• Short-term memory – what it does, why it matters, and how it presents in the classroom

• Working memory – the system the model builds around, and how it operates under pressure

• Long-term memory – what consolidation really involves, and what supports it

• And forgetting – not as failure, but as something the brain is designed to do

Each post will link back to the model to help make what sits inside it more visible. If the model is a map, these posts are about the terrain & about how it actually feels to move through the space.

References

Mccrea, P. (2024, January 17). A simple model of teaching [Tweet]. Twitter/X. https://twitter.com/PepsMccrea/status/1747685287633086862 Image adapted from original by Oli Cav.

But for children who are new to the school (or even just to a new corridor, a new classroom, a new set of faces) something else is happening. Before they grasp a concept or remember a lesson objective, their brain is trying to answer a more basic question:

Where am I?

Not in the abstract existential sense. Not socially or emotionally, though those questions may also be there. This is spatial. Literal. Biological.

They are building a map.

How the Brain Finds Its Way

There are neurons in the brain that only fire when you're in a particular place. Not a general area, not “in the classroom” or “in the corridor”, but specific anchor-point locations. Near the door. At the far-right corner of the table. One spot on the playground you walked across once, and haven’t yet again.

These neurons, helpfully named place cells, live in the hippocampus. Each one becomes active when you're in a specific region of space - a place field. But these cells aren’t arranged like tiles on a map. Their fields aren’t laid out in neat sequence, and neighbouring cells in the brain don’t correspond to neighbouring places in the world. Instead, the hippocampus encodes space relationally, through patterns of activity across many cells, each contributing a piece of the picture. Together, these patterns form a kind of internal map: not drawn to scale, but precise enough to find your way.

They don’t fire just because you’re there. They fire when the place matters.

And running alongside them is a second system: grid cells, located in the entorhinal cortex. These fire in regular intervals, forming repeating hexagonal patterns, a lattice across space that the brain lays down invisibly, automatically. It’s the kind of geometry you’d expect from an architect or a satellite system.

Except it’s happening inside the head of a Year 3 child, walking into a classroom for the first time.

This process is illustrated in Figure 1, where a single grid cell’s activity maps out a regular hexagonal pattern.

Grid cells provide a metric. Place cells provide meaning. Together, they allow the brain to create internal maps of external environments which are rich, navigable models that let us know where we are, where we’ve been, and how to get back.

And critically, these maps aren’t drawn once. They evolve. According to Moser, Rowland and Moser (2015), place cells don’t just represent where you are, they also encode where you’ve been, and where you’re going. In some cases, the same cell will fire during navigation, then again during rest or sleep, as the experience is replayed and consolidated. Mapping, memory, and movement are all entangled, and part of the same cognitive system.

This isn’t metaphor. It’s not like the brain builds a map.

It builds a map.

Which means that when a child starts at a new school, or enters a new classroom, there’s a quiet, unconscious, storm of neural activity beneath the surface whilst they fiddle with their oversized blazer: place cells firing, grid cells aligning, patterns stabilising into something retrievable. And the hippocampus, more than just a memory centre, is working to locate the child not just in space, but in context. In the environment. In the school.

And it’s doing that in the background, while everything else carries on.

What Attention Does

Spatial maps form quickly…but not instantly.

In familiar spaces, the hippocampus retrieves them automatically. But in new ones, it has work to do. That work depends on attention.

Research shows that place fields become more stable when attention is directed to spatial cues (Kentros et al., 2004). If we’re actively engaged in navigating, the map consolidates. If we’re distracted or passive, it doesn’t.

Attention here isn’t just alertness. It’s selective. Purposeful. It flags what matters.

This is where implicit and explicit systems meet. Spatial memory draws on both: the automatic and the intentional. Some routes become familiar without effort. But the ones that stick, that become stable reference points, are usually the ones the brain has reason to notice.

So when a child is slow to settle, or tired, or seems adrift, we might assume they’re disengaged. And sometimes they are. But sometimes, they’re still building the map. And until that’s clear, everything else becomes that little bit harder.

Spatial Memory is Built, Not Taught

When we talk about wellbeing in schools, we often default to pastoral care, or emotional support. And those are essential. But there’s also a cognitive dimension to wellbeing, and it’s one we don’t always recognise.

And part of feeling safe is knowing where you are - not just emotionally, but cognitively.

What do Year 6 children often worry about before starting a new school? Getting lost.

And in today’s society, they rarely get the chance to. We guide them. We supervise. We track their location and prevent them from straying.

To help children learn the school, we often begin with a tour. A walkthrough of key places. Where the toilets are. Where to line up. Where to eat lunch. It feels right.

But spatial memory isn’t built that way.

Being shown isn’t the same as knowing.

If a child is passively led through the school - no decisions to make, no questions to answer, no cues to notice - the hippocampus might not encode anything lasting. Skeletal maps may form, but without repetition or meaning, they fade.

You only remember the path back if you noticed the route.

So instead of one grand tour, we go back again. We link place to purpose.

We let them get it slightly wrong, then slightly right.

Because like any learning that lasts, spatial memory depends on attention, relevance, and challenge - just enough to require effort.

It’s not “Here is the school”.

It’s “Let’s learn it together”.

What Kindness Looks Like

Which do your pupils need more of this September?

A straight path from A to B? Or one that loops and hesitates?

We tend to prefer the straight line. It looks efficient. But the slower path, to begin with, the one with trial and error, is often how memory works.

Through movement. Through noticing. Through doing it again.

Sometimes, finding your way means taking longer than expected.

So we make appropriate room for it. We give children boundaries and expectations and we also reduce the anxiety around getting things wrong, or getting lost.

Not just patience, but practice.

Not just support, but structure.

Some Children Take Longer

This may also reframe how we interpret those early moments: the child who’s late to a lesson. Who hovers outside the wrong door. Who walks the long way round without quite knowing why.

Sometimes, it’s avoidance. But, not always.

Sometimes it’s a spatial map still under construction. A brain that hasn’t yet tagged the room as significant, or the path as familiar or relevant.

The hippocampus doesn’t operate on the timetable. It can’t be rushed. It maps through relevance and repetition and while it will get there, it needs the chance to.

So being kind isn’t the same as letting things slide. It’s about knowing the huge, unconscious & cognitive demands of what’s going on under the surface.

It’s recognising that part of learning, especially at the start, is quite literally finding your place.

And some children take longer.

Not because they’re not trying. Not because they’re less able. But because the map takes longer to build. Because attention is finite. Because for some, especially those navigating uncertainty elsewhere, that background work takes more time.

So we build it into our systems. Into our expectations. Into our understanding of why children are tired. Into our care.

Because children can only learn where they are, once they know where they are.

And some of them aren’t there yet.

But they’re on their way.

💡 Want more research-informed tools like this?

Subscribe for future posts exploring how the brain supports learning, and what that means for schools.

Checklist: Helping Children Learn the School

Supporting orientation through relevance, repetition, and attention

Before the Tour

• ☐ Keep it small: avoid group tours larger than a form where attention will scatter. Could prepped older children take around smaller groups?

• ☐ Clarify purpose: tell the child why you’re showing them each location (“This is where you’ll queue for lunch every day”)

• ☐ Prepare prompts: have open questions ready (e.g. “Which route would you take from here?”) to support engagement

• ☐ Prepare the salient points: there’s no need to show the children where the Head’s office is on their first day.

During the Tour

• ☐ Go slowly: allow time for noticing, not just moving

• ☐ Use anchors: point out key landmarks (“You’ll pass the portacabins on your way to PE”)

• ☐ Let them lead part of the route, even if they get it wrong

• ☐ Repeat names - of places, adults, and routines (“This is Mr Khan’s classroom, where you’ll have Science”)

• ☐ Loop back, revisit a few rooms in a different order to support consolidation

• ☐ Be explicit about uncertainty - model what to do if they’re unsure (“If you’re ever lost, come back to Reception”)

• ☐ Check for listening, check for understanding - repeatedly, using cold-calling

After the Tour (First Week)

• ☐ Retrace the route with the child at key points (e.g. before first break, after lunch)

• ☐ Build in low-stakes recall - ask them to tell you how to get somewhere ("Using your maps, can you show me the way to the Art room?")

• ☐ Check for gaps - gently check if they know how to get to toilets, lockers, or lunch independently

• ☐ Name the feeling - acknowledge that feeling lost is usual (“It takes time to remember all the routes - that’s okay”)

• ☐ Praise orientation, not just punctuality (“You found your way brilliantly, even when the corridor was busy”)

Optional Extras

• ☐ Create a visual map or timetable with photos of key spaces

• ☐ Pair with a peer buddy…but brief the buddy too: focus on repetition, not just companionship

• ☐ Make spatial reflection part of Form Time (“What places do you know now that you didn’t on Monday?”)

Subscribe now

References

Moser, M. B., Rowland, D. C., & Moser, E. I. (2015). Place cells, grid cells, and memory. Cold Spring Harbor perspectives in biology, 7(2), a021808. https://doi.org/10.1101/cshperspect.a021808

Kentros, C. G., Agnihotri, N. T., Streater, S., Hawkins, R. D., & Kandel, E. R. (2004). Increased attention to spatial context increases both place field stability and spatial memory. Neuron, 42(2), 283–295. https://doi.org/10.1016/s0896-6273(04)00192-8

We quiz it, test it, retrieve it.

We use models: working memory, cognitive load, schema theory.

But we rarely stop to ask:

What is memory, really?

Not psychologically.

Not pedagogically.

But biologically.

What actually happens in the brain when we learn something and remember it?

That’s the question Eric Kandel set out to answer, documenting his search in his Ronseal-titled book, In Search of Memory. It's a brilliant book and this post barely skims its surface. What Kandel and his teams uncovered, through decades of molecular neuroscience, is more concrete than most models acknowledge:

Memory isn’t just an arrow in a diagram.

A label in a schema.

It’s a physical change in the brain, etched into cells, shaped by experience.

A note on Kandel: Memory as a personal quest

Eric Kandel was born in Vienna in 1929. As a Jewish child, he lived through the horrific annexation of Austria by Nazi Germany, and his family was forced to flee. What followed was a dislocation…geographic, cultural, and psychological…that would shape the rest of his life.

In In Search of Memory, Kandel’s account weaves together personal history and scientific inquiry, suggesting that memory, for him, is not only a subject of research but also a means of making sense of experience. It seems existential.

“Memory is the glue that holds our mental life together. Without its unifying power, both our conscious and unconscious life would be broken into as many fragments as there are seconds in the day. Our life would be empty and meaningless.”

Kandel, Dudai & Mayford (2014)

Kandel began his academic career in psychiatry, drawn to Freud, but later turned to neuroscience. His goal was radical, particularly for the time: to understand how a lived experience could become biologically embedded.

The sea slug that changed neuroscience

To study memory, Kandel needed a subject that could learn but that wasn’t too complex to trace at the cellular level.

Enter Aplysia, a sea slug with just 20,000 neurons (compared to our ~86 billion), each large and accessible enough to study in detail.

It became the unlikely hero of some of the most important discoveries in modern neuroscience.

Pour one out for Aplysia.

Short-term vs long-term: Function vs structure

So, Kandel’s work began with Aplysia, a sea slug with big beautiful neurons. His early research showed that short-term memory doesn’t require structural change. It simply tweaks how existing synapses function often by modulating neurotransmitter release.

This allows the brain to adjust responses quickly and temporarily.

But that alone couldn’t explain long-term memory, or why, as Kandel likes to joke, he still remembered his housemaid from his childhood.

To persist for days, weeks, or years, memory had to be embedded more deeply.

Synapses didn’t just fire more.

They changed shape.

“Short-term memory produces a change in the function of the synapse…

Long-term memory involves anatomical changes.”

Kandel, 2006

In Figure 1, you can see what that looks like. Repeated activation (just five pulses of serotonin in Kandel’s experiments, or shocks to the poor Aplysia, but that's difficult to diagram) triggers the creation of new proteins, new terminals, and eventually, new connections.

This is the difference between forgetting and remembering.

I would like to remind at this point that Kandel was working with sea slug neurones, and often just three of them in a dish. This is memory in its most reductionist form.

What he was observing was a reflex. Not someone grasping differential equations or making meaning across domains, or learning to read.

So no - drilling something five times doesn’t guarantee learning. And a traumatic event doesn’t have to be viewed 5 times for it to be remembered for life.

Learning is layered. It’s contextual. It’s emotional. It’s social.

Neuroscience shows us what’s possible. Education decides what’s meaningful.

But here we can see, in its most reductionist form, that practice - spaced, focused repetition - makes permanence (not necessarily perfection) possible. But how?

Kandel traced it to the molecular machinery inside neurons. With each repeated stimulus, signals build toward permanence. Memory, quite literally, is etched into the structure of our cells.

It’s not romantic, but that new terminal?

That’s the physical trace of a memory.

Cute.

What actually happens when we form a memory?

Let’s look closer.

Left side: Short-term memory

One stimulus triggers serotonin release. This activates cyclic AMP, which turns on protein kinase A. The synapse becomes more efficient - more neurotransmitter, stronger signal.

But: nothing grows. No structure changes. It fades.

Right side: Long-term memory

With repeated stimulation, the signal enters a new phase. The neuron is serving full boss battle and phase two kicks in.

Protein kinases move into the nucleus of the neuron and trigger gene expression. CREB-1 is activated; CREB-2 is suppressed. Genes are switched on, and the neuron grows new terminals - actual structural change.

Short-term memory is like adjusting the volume.

Long-term memory is like installing a whole new sound system.

Why the brain isn’t a muscle (and why that matters)

At this point, you would be forgiven for thinking: hang on….memory builds with practice….this sounds like a muscle. Muscles…they get trained in a gym. What about if we built a brain…gy…

STOP! NO!

PUT THE CLIPART DOWN!

STEP AWAY FROM THE SMILING BRAIN LIFTING WEIGHTS!

Phew. Thank you.

The brain isn’t a muscle.

It doesn’t grow stronger with generic use. It doesn’t “bulk up” with mental reps. And it doesn’t benefit from vague exercises.

The muscle metaphor has spawned a wave of pseudoscience - Brain Gym, “brain-training”, learning styles - all rooted in the mistaken idea that we can develop cognitive strength through generalised exercises.

But memory doesn’t work like that.

It’s not general. It’s local. It’s structural.

Neurons change in specific circuits, triggered by attention, emotional relevance, and repeated activation. Synapses grow, or retract, based on meaningful use, not rote movement.

Muscles respond to load. Brains respond to meaning.

That distinction matters. Because the myths we teach shape the strategies we choose. And if we want to build memory that lasts, we need to understand how the brain actually works…not how we wish it did.

What educators can take from this

When we talk about memory in the classroom - about retrieval, overlearning, or consolidation - we’re not just building knowledge.

We’re helping create physical change.

Each well-targeted question.

Each spaced review.

Each scaffolded connection.

It’s not just pedagogical technique.

It’s molecular architecture.

That should change how we think about time, depth, feedback, and care. Because learning doesn’t just pass through the mind.

It builds it.

Subscribe now

References

Kandel, E.R., Dudai, Y., & Mayford, M.R. (2014). The molecular and systems biology of memory. Cell, 157(1), 163–186. https://doi.org/10.1016/j.cell.2014.03.001

Kandel, E.R. (2006). In Search of Memory: The Emergence of a New Science of Mind. W.W. Norton & Company.