Subject specialists in primary school

Why can four teachers who each know one subject outteach one who knows the child all day?

A primary school can be staffed two ways. One teacher takes a class for everything: reading, arithmetic, science, art, the wet lunchtime, the fallings-out. Or four teachers each take one subject and rotate through four classes, so every child sees four adults a day and no adult sees a child for more than an hour.

The generalist arrangement has one advantage that is hard to argue with — she knows the child. So the case for splitting the work has to be strong enough to outweigh that, and it is worth asking where its strength comes from rather than assuming specialisation is better because it is what secondary schools and hospitals do.

Ask first what a teacher's day actually contains, and which parts of it are per subject rather than per lesson.

Learning to teach fractions well is not the same work as teaching a fractions lesson. It means knowing the four or five ways children misunderstand a fraction, holding a stock of examples that expose each one, and recognising a wrong answer's cause from its shape. That is a substantial fixed cost, paid once per subject, not once per lesson.

Now count the repetitions. The generalist pays that cost across every subject she teaches, and then delivers each lesson once, to one class. The specialist pays it in one subject and delivers that lesson four times, to four classes. Same investment, four times the use — and because she teaches it four times a day for years, she also gets four times the feedback about which explanations land. This is Adam Smith's observation about the pin factory in 1776, transplanted: the gain is not that a specialist is a better person but that a fixed cost gets spread over more output, and that repetition converts into skill.

There is a second effect, less flattering but real. Primary generalists are not equally strong across subjects. Many entered teaching from an arts or humanities background, and self-reported confidence in teaching mathematics and science runs consistently lower than in literacy. Under a generalist model, that unevenness reaches every child in the class. Under specialisation, each subject is taught by whoever is strongest in it — so the school's weakest subject teaching simply stops happening.

So the case is good. Why, then, does it so often disappoint in practice?

Because we counted only one side. Splitting work has a cost that scales with what has to cross the joins. In a pin factory almost nothing does — a half-made pin is fully described by its own shape, so a worker passing it on carries no information. But what passes between a child's four teachers is not a half-made pin. It is this child is struggling because her parents separated in March, and she goes quiet rather than asking. That is not written on the work. It lives in the sustained attention of one adult over months, and it does not survive a handoff — if anything, it is never formed at all, because nobody accumulates enough hours to form it.

So specialisation pays when the interfaces are cheap and the per-subject investment is high. In primary school the second condition holds and the first badly does not. And the interface costs are not only informational: four transitions a day cost minutes of settling each, and a child with no one adult responsible for her whole day is easier to lose track of.

So the honest answer to "can four specialists outteach one generalist" is: they can, and the mechanism is genuine, but not automatically — it depends on whether the depth gained exceeds the continuity lost, and in the youngest years that trade is often unfavourable. This is where it pays to look at evidence rather than at the reasoning alone. A randomised experiment in Houston elementary schools, published in 2018 by Roland Fryer, assigned teachers to specialised rather than general assignments and found achievement fell modestly, with the effect worst for students with special educational needs — the group most dependent on one adult knowing them. That is the interface cost showing up in the numbers, and it is a useful corrective to an argument that sounds airtight on paper.

Which suggests the real design question is not whether to specialise but where the joins go. Splitting off music, languages and PE — subjects with high specific demands and little dependence on the rest of the day — captures most of the depth gain while cutting almost nothing that needs to be continuous. Splitting literacy from the child who is learning to read cuts precisely the thing that had to stay whole.

Three refinements.

First, the gain from specialisation is real even where the arrangement fails overall. The Houston result does not show that deeper subject knowledge is worthless; it shows that in that setting the continuity lost was worth more. Those are different claims, and conflating them would leave you unable to explain why secondary schools specialise successfully.

Second, age is doing much of the work. Older students carry their own thread between lessons, tolerate transitions, and can seek out the right adult. The younger the child, the more of the continuity has to be supplied by the institution, and the more expensive every join becomes.

Third, this is not an argument that a generalist knows enough. It is an argument that the fix for thin subject knowledge need not be structural: coaching, shared schemes of work, and strong curriculum materials spread the same fixed cost without moving the joins.

• Why music, languages and PE are the subjects almost every primary school specialises first.
• Pedagogical content knowledge: the subject-specific teaching skill that repetition builds.
• How looping — keeping one teacher with a class for two years — attacks the same problem from the other side.

Every term the collection defines is gathered in the glossary.