The Average Child Does Not Exist — Riverside Public School, Coimbatore

In 1950 the United States Air Force had a problem it could not explain. Its pilots were crashing — not in combat, and not rarely. On one bad day, seventeen crashed. The aircraft checked out. The training checked out. Suspicion eventually fell on the cockpit itself, which had been built in 1926 to fit the average pilot.

A young lieutenant named Gilbert Daniels was given the job of re-measuring. He took ten physical dimensions from 4,063 pilots — height, chest, arm length, and so on — and asked a simple question: how many of these men were average on all ten?

He expected most. The answer was zero. Not one pilot in four thousand was close to average on all ten dimensions. Design for the average pilot and you design for nobody.

The Air Force’s response was not to build four thousand cockpits. It was to make the seat, the pedals, and the straps adjustable — the origin of a feature now so ordinary that you adjusted one this morning without thinking.

That is the whole argument for personalised learning, and it was settled seventy-five years ago in a hangar rather than a classroom.

What personalised learning is not

Three misunderstandings do most of the damage, and it is worth clearing them before we go further — not least because two of them get sold to parents.

It is not “learning styles.” The idea that a child is a visual, auditory or kinaesthetic learner, and should be taught accordingly, is the most popular idea in education that the evidence does not support. Pashler and colleagues reviewed it in Psychological Science in the Public Interest in 2008 and found that the studies capable of testing the claim overwhelmingly failed to support it. Children do have preferences. Teaching to those preferences does not improve learning. A diagram helps with things best shown; a sentence helps with things best said. That is a property of the content, not of the child.

It is not a child alone with a screen. Adaptive software can be a useful instrument — it is very good at spacing practice and at noticing which question types a child fails. It is not a teacher, and the evidence for software-led personalisation is far weaker than its marketing. Used well it hands a teacher better information. Used badly it is a worksheet that costs more.

It is not letting a child avoid what they find hard. This is the most damaging version, because it wears the language of kindness. A child who finds long division difficult needs more long division, better sequenced and better supported — not a pathway around it. Personalisation adjusts the route and the pace. It does not move the destination.

The idea underneath it

The theory here is constructivism, and its core claim is modest: children do not receive understanding, they build it. New information is fitted into what a child already knows. Where it fits, it sticks. Where it contradicts, something has to give.

This is why two children can sit through the identical, excellent lesson and leave with different understandings. They did not arrive with the same scaffolding to hang it on.

Lev Vygotsky gave the working version of this in Mind in Society. He described a zone of proximal development: the gap between what a child can do alone and what they can do with guidance. Below the zone, work is easy and nothing changes. Above it, the child is lost. Inside it, with a teacher present, a child does today with help what they will do unaided next month.

Personalisation, stripped of jargon, is the attempt to keep more children inside that zone more of the time.

One honest caveat, because constructivism gets stretched further than it can bear. “Children construct understanding” is a claim about learning. It is not a licence for unguided teaching. Kirschner, Sweller and Clark made this case forcefully in 2006: minimal-guidance approaches — discovery learning, pure inquiry — consistently underperform explicit instruction, particularly for novices, who lack the background knowledge that makes discovery possible. A child cannot discover what they have no means of discovering. The right reading is that instruction should be explicit and pitched at the right place for that child. Not one or the other.

How a school sees a gap early

Here is the part that matters most to parents, and the part most schools describe least clearly.

The traditional rhythm of school assessment is slow. Work is taught for eight weeks, an examination is set, papers are marked, and a mark comes home. By the time you learn your child did not understand equivalent fractions, the class is three topics past it and the gap has begun to compound — because fractions are load-bearing for a great deal that follows.

Formative assessment inverts that. The idea is old and the evidence is unusually strong: Paul Black and Dylan Wiliam’s 1998 review Inside the Black Box remains one of the most cited findings in the field. The mechanism is simple — assessment is frequent, low-stakes, and its purpose is to change the next lesson rather than to produce a mark.

In practice that looks like:

Hinge questions. A single, carefully written question mid-lesson that every child answers at once. It is designed so that each wrong answer reveals a different misconception. The teacher does not learn “sixty per cent got it”; they learn which children think you add denominators, and which think the larger denominator means the larger fraction. Those are different problems needing different responses.
Error analysis rather than error counting. Two children score 6 out of 10 on the same exercise. One has a fragile method and made arithmetic slips. The other has a confident, wrong method and applied it consistently. The mark is identical; almost nothing else is.
Retrieval records. Because material is revisited after a delay rather than tested once and abandoned, a teacher sees not only what a child learned but what they have retained — which is the only version that matters in March.
Reading fluency checks in the younger years. Comprehension problems in Grade 4 are frequently decoding problems that went unnoticed in Grade 1. This is among the highest-value early diagnoses a primary school makes.

Learning analytics is simply what happens when this information is kept rather than discarded. Not a dashboard of vanity metrics — a record, over terms, of which objectives a child has met, which have been re-taught, and which keep reappearing. Patterns become visible at that scale that no single teacher would catch in a single term: the child who is fine in every topic that does not involve multi-step word problems, or the one whose accuracy collapses only under time pressure.

None of this requires expensive systems. It requires small classes, teachers who know the children in front of them, and the discipline to write things down.

What progress looks like without ranking

Ask most parents how their child is doing and the answer arrives as a position: near the top, somewhere in the middle. Rank is the metric schools have trained us all to use, and it is close to useless for the one thing a parent actually wants to know — is my child learning, and what should we do next?

Rank is a poor instrument for three reasons.

It moves when the cohort moves. A child can learn more this term than last and fall five places, because the class improved faster. The number went down; the child went forward.

It compresses everything into one dimension. Daniels’ pilots were not average on all ten measurements, and children are not either. A single position hides the child who is two years ahead in reading and struggling with number bonds — precisely the profile that most needs a specific response.

It tells you nothing you can act on. “Twelfth of thirty” implies no next step. Nobody has ever helped a child by knowing their rank.

The alternatives are more useful and less comfortable, because they are specific:

Growth against the child’s own baseline. Is this child further along than they were in June? Against their own trajectory, not the neighbour’s.
Mastery of named objectives. Not “good at maths” but “can add fractions with unlike denominators; cannot yet do so when the answer needs simplifying.” That sentence tells you what this week is for.
Work samples over time. The single most persuasive evidence of progress is the child’s own book from September beside the same child’s book from January. No scale required.
A named next step. Every useful report says what the child will work on next, and what that will look like when it is achieved.

This is harder to summarise at a family gathering than a rank. It is far more useful on a Tuesday evening when you are trying to help.

A word on comparison, since removing rank does not remove the instinct. Some comparison is legitimate — you are entitled to know whether your child is where children of that age generally are, and a school that cannot answer that is hiding. The distinction is between calibration (is this child broadly on track for their age?) which is useful once or twice a year, and ranking (who is beating whom?) which is a standing invitation for a child to conclude that ability is fixed and theirs is settled. That conclusion, as we wrote in our note on five books for parents, is the single most expensive belief a child can hold.

Five questions worth asking at the next parent meeting

What specifically can my child do now that they could not do in June?
What is the next thing they will learn, and how will we know when they have?
Where did they struggle this term, and how did you find out — from an exam, or earlier?
Is there anything you are watching but not yet worried about?
What can we do at home that would actually help, rather than just add pressure?

A school that can answer all five without reaching for a mark sheet is doing this properly.

Where we stand

We should be plain about something. Personalised learning is easy to promise and expensive to deliver, and the expense is not technology. It is class size and teacher time. A teacher with sixty children can identify a struggling child; they cannot respond to fifteen different misconceptions in a forty-minute lesson. This is the entire reason we hold classes small — not as a comfort, but because it is the precondition for everything described above.

Benjamin Bloom put the ceiling on this in 1984 with what he called the two-sigma problem: children taught one-to-one dramatically outperformed those in conventional classes. The original magnitude has not held up under later scrutiny, and one-to-one tutoring for every child is not a thing any school can offer. But the direction of the finding has never been in doubt, and it defines the work: to close as much of that distance as a real classroom allows.

Adjustable seats, not four thousand cockpits.

If you would like to see how this works in practice, come and sit in on a lesson. It is more convincing than anything we can write.