[This is the sixteenth of seventeen finalists in the book review contest. It’s not by me - it’s by an ACX reader who will remain anonymous until after voting is done, to prevent their identity from influencing your decisions. This entry was promoted to finalist status by readers; thanks to everyone who voted! - SA]


Why are all children so bad at learning in school?

Seriously, they’re terrible at it, and nobody ever calls them out as a group. We call out individual children as failing. We call out individual schools and school systems as bad. But the much more dramatic contrast is between learning in school and learning in any other context.

In their first five years, kids learn to understand 25,000 words, even if nobody is actively helping them, at the same time as they’re learning most of what they’ll ever know about physics, psychology, and how to pilot a human body. They then struggle to match this vocabulary acquisition rate over their next ten years, despite expert attention, a wealth of resources, personal encouragement, and even prizes.

After weeks of trying, my teacher gave up on getting me to correctly label the countries and capitals in a map of South America. Yet I quickly learned to navigate the New York City Subway.

Through involuntary cultural osmosis, I could probably pass a test on the characters, plot, and setting of Twilight, despite having never read any of the books or watched any of the movies. Yet there are books I read in school (good books, written to be enjoyed!) where I now couldn’t tell you the main character’s name.

(I was a straight-A student, by the way. Everybody fails at least this hard.)

As an adult, when I’m in the middle of researching something, and I get hungry, I get up, go to the kitchen, get a snack, come back, and keep researching. In most classrooms, children are absolutely forbidden from leaving to go get a snack because they would always claim to get hungry and they would never come back.

Which is not to say that children don’t try, or don’t care. They sweat and cry over tests. They care so deeply that they have nightmares about missing class that last well into adulthood. But by any standard other than comparing them to other schoolchildren, they universally fail. Why?


In 1958, John Holt was trying to teach math, writing, and French to a class of ten-year-olds. It was going as this normally does–a large proportion were absolutely terrible at it. He wrote to a colleague,

If you live at a small school, seeing students in class, in the dorms, in their private lives, at their recreations, sports, and manual work, you can’t escape the conclusion that some people are much smarter part of the time than they are at other times. Why? Why should a boy or girl, who under some circumstances is witty, observant, imaginative, analytical, in a word, intelligent , come into the class-room and, as if by magic, turn into a complete dolt?

He collected the memos and essays he wrote that year into a book, How Children Fail, meant to be read by educators and parents to help them understand what is going wrong in schools, as a first step toward figuring out ways of fixing it. Holt was a big believer in “hold off on proposing solutions until you’ve really analyzed the problem.” When I first read this book, though, I wasn’t in its intended audience. I was nine years old, which meant it was a handbook, not on education, but on rationality.


More than thirty years after those students were frustrating Holt, I was the dolt in another small, private school. At home I was teaching myself to code in HyperTalk. At school I was exactly this kid from a Holt memo:

…his school papers are as torn, smudged, rumpled, and illegible as any I have ever seen. The other day the class was cleaning out desks, and I was “helping” him. We got about a ream of loose papers out of the desk, and I asked him to put them in the notebook. As always, when he is under tension, his face began to get red. He squirmed and fidgeted, and began to mutter. “They won’t fit, the notebook’s the wrong size”– which wasn’t true. Finally he assembled a thick stack of papers and began to try to jam them onto one of the rings in his notebook, not noticing that the holes in the papers were at least a half-inch from the ring. As he pushed and fumbled and muttered, I felt my blood pressure rising until, exasperated almost to rage, I said loudly, “For heaven’s sake, leave it alone, do it later, I can’t stand to watch any more of it!”

My school held an annual fundraiser–parents donated used items, and the school sold them to other parents. There was always a pavilion full of unorganized boxes of old books. At the end of the day, students were allowed to take home any unsold books. We were told that any we didn’t take would be destroyed, sent to a landfill. This was probably true (it viscerally horrifies me to this day), but if it was a lie it was a really clever tactic. It certainly motivated me to pick out books by the dozen that were out of my comfort zone–if they ended up unread on my shelf, that’d be no worse a fate than otherwise awaited them. It was a great engine of serendipity, and my reading this book was its best product–a book that no adult would have bought for me and that I probably wouldn’t have ever checked out from a library.

Holt describes the bad habits of thought that schoolchildren fall into, many of which will be familiar to readers of 21st-century rationality texts. On positive bias,

Sometimes we try to track down a number with Twenty Questions. … They still cling stubbornly to the idea that the only good answer is a yes answer. This, of course, is the result of the miseducation in which “right answers” are the only ones that pay off. They have not learned how to learn from a mistake, or even that learning from mistakes is possible. If they say, “Is the number between 5,000 and 10,000?” and I say yes, they cheer; if I say no, they groan, even though they get exactly the same amount of information in either case. The more anxious ones will, over and over again, ask questions that have already been answered, just for the satisfaction of hearing a yes.

On failing to think concretely or across domains,

A friend was studying for a chemistry test. He was trying to memorize which of a list of salts were soluble in water. Going through the list, he said that calcium carbonate was soluble. I asked him to name some common materials made of calcium carbonate. He named limestone, granite, and marble. I asked, “Do you often see these things dissolving in the rain?” He had never thought of that. Between what he was studying for chemistry and the real world, the world of his senses and common sense, there was no connection.

And he makes similar claims, similarly argued, to those of Paul Graham and Eliezer Yudkowsky, that the strategies that lead to nominal success in school are often the ones that stop at superficial understanding of the subject–hacks to be able to get to the correct answer quickly, without ever really looking at the problem.

These lessons were helpful to me at the time. As someone generally much better than peers at getting the right answer on a test, I’d been eager to believe that that’s what merit in school was. Holt taught me that I could get more out of class if I let myself be, nominally, a worse student: one who avoided memorization, who questioned what was being taught, who reached for generalizations beyond what would help with the next test. And for the areas where I already knew I was failing, Holt helped me a little to cope with the shame, anxiety, and avoidance spirals that grew out of and perpetuated that failure. Eventually I was able to use a 3-ring binder.

Per the last demographic survey of the readership of this blog, you are most likely not nine years old. However, you are almost certainly a former nine-year-old, and that’s another excellent audience for this book. Holt, like Graham and Yudkowsky, sees school as instilling permanent cognitive biases–habits that are best unlearned whenever you can.


Holt identifies three broad categories of failure modes.

Strategy - Perhaps uniquely for a rationality text, Holt identifies thinking strategically as a failure. Not in general–he encourages adults to ask themselves more often where they are trying to get, and then whether their current approach is getting them there. And he taught, often, through strategy games. But for a child, learning due to disinterested curiosity is much more effective than learning due to incentives. We’re born with powerful drives to learn (learning could hardly start as a learned behavior). Replacing that with a desire to gain reward or approval, or to avoid punishment, or to even to Be A Good Student, distorts behavior. Children, consciously and unconsciously, start trying to maximize their perceived score. They seize on the most reliable way of figuring out the right answer, even if that’s something overfitted to the classroom, like reading the teacher’s face. They find a level of achievement they can reliably hit, then manage expectations to make sure they’re never pushed to a higher one. If they’re punished for giving wrong answers in class, they’ll stop talking and disengage when they think they might not understand something. It’s a familiar problem across domains–once you attach incentives to a convenient measurement of something, it stops being a good measurement. Readers of this blog may remember examples from machine learning or centrally planned economies. Holt’s revised edition gives a quick one from the Navy:

I remember an old chief machinist on an obsolete training submarine in Key West saying bitterly about his worn-out engines, which he had spent many hours polishing up for an official inspection, “They shine, don’t they? Who the hell cares if they don’t work?”

Fear - The more visibly dysfunctional response to incentives. Fear is the mindkiller, as another U.S. Navy veteran would write in a book published the year after Holt’s. A student afraid of failure cannot acknowledge their own mistakes, and therefore cannot learn from them. Students labeled “Gifted” are so terrified of losing that label that they panic when they encounter something they’re not excellent at, end up doing even worse, and try to avoid that whole area whenever possible. Holt describes a lot of his students as “emotionally incapable of checking their work”–not as a rational response to any incentive, but because looking for mistakes is like checking under the bed for monsters.

Boredom - Holt talks at length about children being bored in schools, but I’m going to skip over that for the same reason people rarely yawn in plays–you don’t want it to spread to the audience. I don’t think I need to sell you on the notion, anyway.


All of these failure modes have the same cause. Learning in schools is externally imposed, not intrinsically motivated. School requires a power structure, with incentives and accountability, to cause children to reliably learn, to learn the Right Things, and to be able to prove that they did. In 1998, you didn’t need to assign chapters to get one out of three kids to read Harry Potter and the Chamber of Secrets --they read it because they wanted to see what happened next. You didn’t need to test them in order to find out how much they’d read and understood–they would insist on telling you about the book’s plot, and ask questions about anything they didn’t understand. You didn’t need to find any way to get them to reliably sit still for it–they could be allowed to set their own schedule.

But say you wanted every kid who could to read and understand it. Then you’d need a school.

Similarly, you really don’t need a school to get a decent percentage of children to learn math. I was as ravenous for extracurricular algebra as I was for extracurricular Harry Potter. But to get Scott Alexander to learn any not-directly-useful math, you seem to need a school. (If you replace “math” with “maps”, our positions are reversed.)

The basic concept of school, then, requires incentives, fear, and boredom. And the fear component really isn’t optional, says Holt, speaking from experience. In one of his memos, he wrote

The other day I decided to talk to the other section about what happens when you don’t understand what is going on. We had been chatting about something or other, and everyone seemed in a relaxed frame of mind, so I said, “You know, there’s something I’m curious about, and I wonder if you’d tell me.” They said, “What?” I said, “What do you think, what goes through your mind, when the teacher asks you a question and you don’t know the answer?” It was a bombshell. Instantly a paralyzed silence fell on the room. Everyone stared at me with what I have learned to recognize as a tense expression. For a long time there wasn’t a sound. Finally Ben, who is bolder than most, broke the tension, and also answered my question, by saying in a loud voice, “Gulp!” He spoke for everyone. They all began to clamor, and all said the same thing, that when the teacher asked them a question and they didn’t know the answer they were scared half to death. I was flabbergasted–to find this in a school which people think of as progressive; which does its best not to put pressure on little children; which does not give marks in the lower grades; which tries to keep children from feeling that they’re in some kind of race. I asked them why they felt gulpish. They said they were afraid of failing, afraid of being kept back, afraid of being called stupid, afraid of feeling themselves stupid. Stupid. Why is it such a deadly insult to these children, almost the worst thing they can think of to call each other? Where do they learn this? Even in the kindest and gentlest of schools, children are afraid, many of them a great deal of the time, some of them almost all the time. This is a hard fact of life to deal with. What can we do about it?

Nothing, per the conclusion of the book:

The idea of painless, non-threatening coercion is an illusion. Fear is the inseparable companion of coercion, and its inescapable consequence. If you think it your duty to make children do what you want, whether they will or not, then it follows inexorably that you must make them afraid of what will happen to them if they don’t do what you want. You can do this in the old-fashioned way, openly and avowedly, with the threat of harsh words, infringement of liberty, or physical punishment. Or you can do it in the modern way, subtly, smoothly, quietly, by withholding the acceptance and approval which you and others have trained the children to depend on; or by making them feel that some retribution awaits them in the future, too vague to imagine but too implacable to escape. You can, as many skilled teachers do, learn to tap with a word, a gesture, a look, even a smile, the great reservoir of fear, shame, and guilt that today’s children carry around inside them. Or you can simply let your own fears about what will happen to you if the children don’t do what you want, reach out and infect them. Thus the children will feel more and more that life is full of dangers from which only the goodwill of adults like you can protect them, and that this goodwill is perishable and must be earned anew each day.


Holt never stopped looking for ways to make school incrementally better, but the basic tragedy seems baked into the very concept. We have things we want all children to learn, so we need institutions to teach them. Children on their own are curious, but not driven to learn exactly those things, so we need ways to encourage them. Institutions need accountability, so we need ways to prove that the children are learning those things. So we impose incentives and legible performance metrics, and that kind of breaks kids. And we tend to stay broken for the rest of our lives.

Holt wasn’t always explicit about this, at least until the revised edition. He was still employed by schools, still trying to make schools work, and therefore reluctant to characterize schools as outright destructive. But not that reluctant, and the connections are inescapable. Failing students blindly follow “recipes” to solve math problems, too insecure to stop and think about why the recipes exist or whether they’re moving in the right direction. Failing teachers use traditional pedagogy even when it’s demonstrably not working, too insecure to even consider they might be making things worse. Fear and insecurity beget fear and insecurity down the generations, with bullied students growing up to become bullying teachers. Students conditioned to think of schoolwork as an abstract character-building exercise become parents who don’t care whether what’s taught in school has any meaning or utility. I suspect this is at least a little true and is why it took a frustratingly long time for me to persuade my parents to let me drop out of school. (I ended up “self-schooling” for three years, from 15 to 18, before going to a traditional four-year liberal arts college. It was great.)


The long-term psychological effects of school are, as far as I know, ambiguous. Perhaps it damages our rationality or disinterested love of learning. Perhaps that’s just a natural part of growing up, and schools help us develop the skills that we’ll need to do work we’re not always interested in. The child welfare case against directed learning is also hardly a slam dunk. Plenty of children like the control structure in school; it’s gamifying something that’s already fun. But if we set those issues aside for a moment and zoom out, there seems to be a clear tradeoff between efficiency and control. Children are naturally learning at high velocity, but may not be learning the most useful or important things they could be. You can exert control to optimize the set of topics they’re learning, but at a cost of reduced velocity.

Holt gives an anecdote of a fifth grader caught sneakily reading a science book when he was supposed to be learning about “Romans in Britain.” By forcing him to put the book away, the teacher traded an hour of high quality science education for an hour of low-quality history education, during which the child is less engaged and will remember less. But perhaps the marginal value of the latter was still higher; maybe knowing one thing about Romans in Britain, rather than zero, will be more valuable than knowing a hundred more things about a scientific topic you’ve already gone deep on. When you take away the babysitting and socialization benefits of school (as has happened recently) this marginal value proposition is what’s left. You might not buy it for the Science vs. Romans story, but surely forcing kids to learn just a little reading, writing, and arithmetic is worth it in the long run, no matter what else they’d rather be learning. But we’ve taken this to an absurd extreme, where self-directed learning becomes a tiny part of a typical childhood, something you carve out a little time for during your nth year in state-mandated courses of study you have no interest in. Why did we all have to learn that the mitochondria is the powerhouse of the cell? That’s not even the most interesting or useful fact about mitochondria. Society does not depend on universal lifelong literacy in cell biology; we could get along just fine if 5% of us knew the powerhouse thing, and the other 95% had skipped bio class that year. I assert we would, in fact, be better off, because those 95% would have learned more. Our desire for scalable, testable instruction leads, by middle school, not to ensuring breadth of learning, but to forcing identical amounts of depth in the same set of somewhat arbitrary subjects.


How Children Fail was published sixty years ago, when the most exciting piece of educational technology was a set of wooden blocks. John Holt wrote a revised edition of How Children Fail in 1982, after becoming an advocate for homeschooling and child emancipation. The original edition ended by calling for reform of the school system, saying it should provide buffet tables for students to browse, rather than force feeding. The revised edition adds a postscript:

[E]xcept in very rare circumstances the idea of special learning places where nothing but learning happens no longer seems to me to make any sense at all. The proper place and best place for children to learn whatever they need or want to know is the place where until very recently almost all children learned it–in the world itself, in the mainstream of adult life. If we put in every community, as we should (perhaps in former school buildings), resource and activity centers, citizens’ clubs, full of spaces for many kinds of things to happen– libraries, music rooms, theaters, sports facilities, workshops, meeting rooms- -these should be open to and used by young and old together. We made a terrible mistake when (with the best of intentions) we separated children from adults and learning from the rest of life, and one of our most urgent tasks is to take down the barriers we have put between them and let them come back together.

He also notes in passing that computers (“though only if very different from present ones”) might be able to help children clarify their thinking. Well. Tragically and frustratingly, Holt died only three years later, and so he didn’t quite get to see computers become that mixed-age community center he’d envisioned. I was a privileged kid, and when self-schooling I had access to all sorts of resources. But all I really needed, back in 2000, was AOL. Technology provides education in attractive snack form and allows special interests to scale.

Screens are an incredibly strong draw to anyone over, say, four or five years old. You can get a four year old passionately interested in learning to read by making literacy tools the only video game available. I don’t know if Holt would fully approve, but you don’t really need a control structure anymore to get kids to study the exact thing you want them to–you just need to make the only sophisticated electronic device readily available be one programmed to teach it. Then, a little later, introduce them to the concept of edutainment videos on YouTube, and for better or for worse, The YouTube Algorithm will take it from there.

The internet’s collapse of scale also removes much of the need for standardized education. Holt talks about a fifth-grader who was fascinated by snakes, wanted to know everything about them, and learned more math when it happened to be incidentally involved in snake facts than in a math class. Snake Kid’s school couldn’t be expected to create a whole Snake Curriculum just for this one student, so if Snake Kid wanted an adult to explain things to him, those things would not be Snake Facts. But today, an endless series of snake videos is just a click away.

And the internet is a far more powerful engine of serendipity than a pavilion of unwanted books.


School is an institution. We come to depend on it, it defends itself, it’s not easy to change from inside or out. I mostly believe the critiques in Holt’s book, but I generally haven’t felt that radical reforms were possible. I’ve encouraged children who don’t need as much socialization or supervision to consider getting parental support for dropping out of school, and that’s been more or less it, reform-wise. But we maybe do have a moment here, in 2021. We’ve had a massive disruption in traditional schooling, which could be an opening to escape from inertia. We’ll at least have a natural experiment as different schools physically reopen at different times.

What would happen if we let kids choose how to allocate more of their time, and gave them the support and resources we could? What if the free daycare provided by state schools wasn’t as coupled to regimented instruction, but still included books, computers, and adults who would help explain things if asked? A lot of kids would spend their time in ways that didn’t look very productive to us–binge watching TV, inventing new sports, making a thousand paper cranes. Almost all of them would end up with educational deficiencies that looked shocking to us. I am skeptical that in this day and age the consequences would be dire, though. It’s really easy to catch up when you need to (did you know you can just look up what the powerhouse of a cell is?). With less Educational Attainment gatekeeping down the road, we could more safely test this. And I really think we should, because while the benefits of schooling are ambiguous, some of the costs are not. Kids are definitely less free to pursue their passions. Unless you have a really exceptional amount of energy and parental support, school puts a ceiling on your personal productivity as a child. And long-term consequences aside, school is just not maximally fun. Your greatest capacity for joy was as a child. The time of a ten-year-old is just as valuable to her as the time of an adult is to the adult. And yet, even when a classroom format isn’t working well for a ten-year-old and they need some kind of other instruction, we often still stash them in a classroom to kill time being miserable and frustrated in between actual study sessions, as if there’s no benefit in them having fun.

Suppose you were asked to spend a day as a fifth-grade student. Six hours, with breaks, getting taught things you mostly already knew, with an adult in the room who will try to catch you not paying attention and at minimum be offended if she does. You might do it for the novelty or nostalgia, I suppose. For a day. How much would someone have to pay you to do it for a week? For a month? How good a case for the social utility of the thing would somebody have to make for you to do it for nine months, for free? I was a lower-case-g-gifted fifth grader in a school without a Gifted program, and when I complained that my time was being wasted this way, nobody cared as much as I thought they should, because nobody thought there was anything else worthwhile I could be doing. I’m glad I had this book back then to tell me they were wrong.