Why we teach the way we do

Picture a classroom. You already have one in your head, and it doesn't matter much which country you grew up in: thirty-ish people the same age, sitting in rows or clusters, facing one adult who is standing up. A board behind the adult. A timetable that chops the day into blocks of forty or fifty minutes. A bell.

Almost none of that is obvious. Almost all of it is an accident of administration.

Grouping children by their year of birth isn't a fact of nature. Neither is teaching six subjects a day in rotation, or stopping for six weeks in summer, or setting work to be done at home in the evening. Each of those has a date and a reason, and once you know the reasons a lot of school stops looking like pedagogy and starts looking like logistics.

Nobody designed it, exactly

You'll often hear that schools were built to produce obedient factory workers. It's a satisfying line and it doesn't hold up well. Mass schooling was pushed hardest by people who wanted literate citizens, soldiers who could read orders, and children who could read scripture — and it was frequently opposed by industrialists, who wanted those children in the mill rather than in a classroom.

What is true is that the shape of school came from administrators solving an administrative problem. Prussia built a state system in the eighteenth and nineteenth centuries with trained teachers, a set curriculum, inspection and compulsory attendance. It worked, and it was copied. Horace Mann went to look at it in 1843 and came back to Massachusetts full of it. So did visitors from France, Japan and much of the United States.

The single most consequential import was age-grading. Sort children by birth year, teach each group a fixed year's worth of material, move them all up together in September. It's easy to staff, easy to inspect and easy to explain to parents. It also assumes every child in a room learns at roughly the same rate, which is the one thing everybody involved knows isn't true.

Before that, the standard arrangement was a single room with every age in it at once, the teacher working through pupils individually or in small groups while everyone else got on with something. That's the schoolhouse in the old photographs. It handled mixed ability effortlessly and it handled scale terribly.

The alternative that nearly won

For a while it looked as though mass education would take a completely different form. Andrew Bell and Joseph Lancaster, working separately around 1800, developed the monitorial system: one paid teacher, hundreds of pupils in a single hall, and a chain of older children — monitors — drilled on a lesson and then sent to drill ten younger ones each.

It was astonishingly cheap, which was the point. Lancasterian schools spread through Britain, the United States and Latin America. A single master could nominally be responsible for five hundred children.

It faded because it was rigid, because it depended on rote drill for everything, and because as states started paying for education properly they could afford real teachers in smaller rooms. But the idea keeps coming back in other clothes. Peer tutoring is monitorial teaching with a better reputation, and the research on it is broadly positive — children explaining material to other children learn it well, partly because explaining forces you to retrieve.

Small technologies that changed the room

The wall-mounted blackboard is usually credited to James Pillans in Edinburgh around 1801, and versions spread quickly through military academies and then everywhere else. It sounds trivial. It wasn't. Before the board, instruction was mostly one-to-one or read aloud from a text, because there was no way to show the same thing to forty people at once. The board is what made whole-class teaching physically possible, and the room reorganised itself around it — rows, all facing one way.

Cheap paper and printing did something similar for what pupils produced. Slates were erasable, so nothing survived a lesson; exercise books meant work could be collected, marked, kept and compared. Marking as we know it is a consequence of paper getting cheap.

Then there's the timetable. The rigid period is younger than you'd think, and in American secondary schools it hardened around something called the Carnegie Unit, defined by the Carnegie Foundation around 1906. It was invented as an accounting device — a way of standardising what "a year of study in a subject" meant, tied up with a pension scheme for college professors, so universities could compare applicants from different high schools. It was a measure of time spent in a seat, not of anything learned. It's still, more or less, how secondary schooling is organised.

The summer holiday isn't agricultural

The story everyone tells is that the long summer break exists so children could bring in the harvest. It's a good story with the seasons wrong.

Harvest in most of the northern temperate world is late summer into autumn, and planting is spring. A calendar designed around farm labour would break in those months, and rural schools did exactly that — they typically ran two short terms, winter and summer, with gaps in spring and autumn when the fields needed hands.

Urban schools had the opposite pattern. Many were open nearly year-round, but attendance collapsed in July and August because cities were unbearable, disease was worse in the heat, and any family that could afford to leave did. When reformers pushed to standardise the school year across cities and countryside in the mid-to-late nineteenth century, the long summer gap was the compromise that stuck. Kenneth Gold's history of the American school calendar walks through the paperwork.

So the holiday is about heat and middle-class travel, not hay.

What the evidence says about class size

This is the reform everyone wants and almost nobody can afford, and the research is more awkward than either side admits.

The strongest evidence comes from Project STAR in Tennessee, which ran from 1985 to 1989 and did the thing educational research almost never manages: it randomly assigned around 11,600 children, and their teachers, to small classes of about 13 to 17, regular classes of 22 to 25, or regular classes with a teaching aide. Random assignment matters enormously, because otherwise you can never tell whether the small class caused the result or the sort of school that runs small classes did.

The small classes did better. The effects were clearest in the earliest years, were larger for disadvantaged children, and some follow-up work found traces still visible years later.

The caveats are what get left out. The reduction that produced those gains was roughly a third of the class, not two or three pupils. Cutting a class from 30 to 27 buys you very little and costs a great deal, because staffing is the largest line in any school budget. And a class-size cut across a whole system means hiring a lot of teachers quickly, which tends to lower the average quality of the teaching workforce and can cancel out the gain. Eric Hanushek has spent decades making that argument and it hasn't been dismissed.

The honest summary: class size is real, but it's an expensive way to buy an effect, and there are cheaper things that buy more.

Homework, and the age split nobody mentions

Harris Cooper has spent a long career synthesising the homework research, and his reviews keep landing in the same place. In secondary school there's a moderate positive relationship between homework and achievement. In primary school it's close to nothing.

That's a fairly striking result to sit on for decades, given that primary homework is where most of the household misery is generated.

There are decent explanations. Younger children have weaker study skills and less capacity to work unsupervised, so home practice is unreliable at best and actively teaches bad habits at worst. Older students can practise properly and benefit from spaced repetition of material. There's also a ceiling — beyond an hour or two a night at secondary level, the relationship flattens and then bends the wrong way.

One thing that does survive the evidence is reading. Children reading, at home, for pleasure, is about as consistently positive as anything in the field. It's just not really homework.

Learning styles don't exist, and everyone teaches as though they do

Let's be blunt about this one, because hedging has done real damage.

The idea is that each of us is a visual, auditory or kinaesthetic learner, and that instruction matched to our style produces better results. The specific claim that matters is called the meshing hypothesis: matching the method to the learner should improve learning relative to mismatching it.

To test it you need to identify learners' styles, randomly assign them to matched or mismatched instruction, and check whether the matched groups did better. Pashler, McDaniel, Rohrer and Bjork reviewed the literature in 2008 for Psychological Science in the Public Interest and found almost no studies that had run the right design at all. Of the few that had, the results didn't support meshing. Later work has kept failing to find it.

Meanwhile, surveys of teachers across many countries routinely find that a large majority believe in learning styles, and there's a whole industry selling questionnaires that sort children into categories.

Two things are true at once, and conflating them is where the confusion lives. People do have preferences — they'll tell you sincerely that they prefer diagrams to talking. Preferences are real. What isn't supported is that teaching to the preference produces more learning. And material has its own demands regardless: you learn geography from maps and pronunciation from listening, whoever you are.

What does work is mostly unglamorous

Cognitive psychology has produced a short list of techniques with unusually solid support, and none of them are exciting enough to sell at a conference.

• Spacing. The same total study time spread over weeks beats the same time in one sitting, by a lot. Hermann Ebbinghaus was documenting this on himself in the 1880s and it's held up ever since.
• Retrieval practice. Testing yourself isn't just measurement, it's the learning event. Roediger and Karpicke showed in 2006 that students who read a passage once and then tested themselves repeatedly retained far more a week later than students who reread it four times — even though the rereaders felt more confident.
• Interleaving. Mixing problem types up rather than doing twenty of one kind in a row. Performance during practice gets worse. Performance later gets better.
• Explaining and elaborating. Making yourself say why something is true, out loud or on paper.

Dunlosky and colleagues rated ten common study techniques in 2013. Practice testing and distributed practice came out on top. Highlighting, rereading and summarising — which is what practically every student actually does — came out near the bottom.

Why the research doesn't reach the classroom

There's no mechanism. That's the short answer, and it's more damning than any individual failing.

Medicine has journals, professional bodies, regulators and continuing education that push findings towards practitioners, imperfectly but constantly. Teaching mostly doesn't. A teacher trained in 2003 may never encounter a research summary again unless they go looking, and the material that reaches them through commercial training is chosen for how well it sells.

Then there's a genuinely nasty psychological obstacle. Robert Bjork calls the useful ones desirable difficulties: the techniques that produce durable learning make the learning feel harder and slower while it's happening. Rereading a chapter feels smooth and productive. Closing the book and trying to recall it feels like failure. Both students and teachers are being asked to trust an outcome they can't perceive against an experience they can, and the experience wins nearly every time.

Structural pressure finishes the job. Curricula are long, exams are fixed, and a teacher who slows down to space and interleave is a teacher who hasn't covered the syllabus. Spacing across a year requires timetabling decisions no individual teacher controls.

The bits schools argue about hardest

Reading instruction has been fought over for a century — whether children should be taught letter-sound correspondences systematically or should absorb reading from meaningful text with light guessing support. Large reviews, including the US National Reading Panel in 2000 and the Rose Review in England in 2006, came down on systematic phonics as a necessary component, particularly for children who don't pick reading up easily. Comprehension needs a great deal more than decoding, but decoding isn't optional, and several school systems have spent the last few years unwinding programmes built on the other assumption.

Examinations are the other permanent argument, and they're much older than the schools. China ran competitive written examinations for the civil service for well over a thousand years — the keju, running in recognisable form from the Tang dynasty until it was abolished in 1905. It was the first large-scale attempt to select on demonstrated ability rather than birth, and it produced everything modern exam systems produce: genuine mobility for a few, cram schools, cheating scandals, and a curriculum bent entirely towards whatever was on the paper.

That last effect has a name now. Goodhart's law — when a measure becomes a target, it stops being a good measure. Any assessment that carries real consequences will reshape teaching around itself, and no amount of insisting that teachers shouldn't teach to the test has ever changed that.

Where that leaves you

None of this makes schools bad. Mass literacy is one of the largest achievements of the last two centuries, and the age-graded classroom delivered it at a scale nothing else has managed. England made elementary schooling available in 1870, compulsory to age ten by 1880 and free by 1891; most of Western Europe was somewhere similar by 1900. Within a few generations reading went from a minority skill to an assumption.

The system just carries an enormous amount of inherited furniture — the birth-year cohort, the fifty-minute block, the summer gap, the evening worksheet — that was settled for reasons which no longer apply, and which is now defended mainly because it's what everyone remembers.

If you take one usable thing from the research, take the testing effect. Shut the book and try to remember it. You'll do worse in the moment and better in a month.