Why Learning Gaps Compound, and How to Catch Them Early

A Grade 9 student cannot solve a two-step linear equation. Her teacher reteaches the method, slowly, twice, and she still cannot do it. The actual break is four years upstream: she has never been secure with equivalent fractions, so every time the algebra requires multiplying through by a denominator she loses the thread and guesses. No amount of reteaching Grade 9 content will reach a Grade 5 gap.

This is the mechanic behind most of what gets recorded as “weak in maths”, and it is almost entirely invisible in the data a school collects, because the data is organised by the term in which the mark was awarded rather than by the concept that failed.

A gap is a tax, not a hole

In hierarchical subjects, later content assumes earlier content is automatic. When a prerequisite is missing, the student does not simply lack one piece; every downstream lesson costs more working memory, because part of their attention is spent reconstructing something that should be free. Less of the new material lands, which creates the next gap, which raises the tax again.

Two years of that and the accumulated effect is indistinguishable from low ability — to the teacher, to the parents, and most damagingly to the student, who has by then formed a stable opinion of herself as someone who is bad at the subject. That is the compounding. It is not the original gap getting larger. It is the interest.

Why the school calendar catches it late

Terminal assessment reports a score after the window for acting has closed. The mark aggregates across every concept in the term, so a middling result can hide a total failure on one idea underneath solid performance on five others. And a report card is a communication instrument, designed to summarise for a parent, not to diagnose for a teacher. None of these are faults of the people involved; they are properties of measuring at the end.

What catching it early actually requires

Items mapped to concepts, and concepts mapped to prerequisites

Every question needs to carry the concept it tests and the concepts that concept assumes. Then a wrong answer points somewhere specific, and a pattern of wrong answers points upstream. Without that mapping, a low score is a fact with no address, and the only available response is to reteach the current topic more loudly.

Short, frequent and low-stakes

Ten minutes each week produces far more usable signal than ninety minutes each term, and it produces it while there is still time to act. Frequency also lowers the stakes of any single measurement, which reduces both the anxiety and the incentive to copy — two things that quietly corrupt a lot of assessment data.

Telling “does not know” apart from “did not read the question”

A consistent wrong answer is more informative than a low score. A student who subtracts numerators and denominators separately, every time, has a specific and correctable misconception. A student who gets four of the same item type right and one wrong has a lapse of attention. Treating those two the same wastes the intervention on the student who did not need it and mislabels the one who did.

Remediation aimed upstream

If the diagnosis is Grade 5 fractions, then the work is Grade 5 fractions. That is easy to say and socially difficult to do: a fourteen-year-old will not accept material that is visibly labelled as four years below her class in front of her peers. It has to run alongside the current topic rather than replacing it, it has to be private, and it has to be short enough to finish. This is where after-hours tutoring does its most useful work, because the embarrassment cost of asking the same question for the fifth time drops to zero.

Somebody with the hours

The reason schools do not do this is not ignorance. It is arithmetic. One teacher, forty students, six periods a day, and a syllabus with a fixed end date. Diagnosing individual prerequisite gaps by hand is simply not available at that ratio. The automatable parts — tagging items, grouping students by shared misconception, ranking gaps by how much downstream content depends on them — are where a platform earns its place in a school, because they give back the one resource that cannot otherwise be created.

Where this argument does not apply

Not every subject is a chain. History, literature and most of the humanities build cumulative skills — argument, evidence handling, close reading — but the content is not strictly sequential, and a student who missed the Industrial Revolution is not thereby blocked from understanding decolonisation. Applying a rigid prerequisite model everywhere produces a mechanical curriculum and a lot of pointless remediation.

Use the model where the dependency is real: arithmetic into algebra, algebra into calculus, phonics into fluency, vocabulary into comprehension, mechanics into electromagnetism, the early programming constructs into everything that follows. In those subjects the chain is not a theory, it is the structure of the discipline.

The signal worth watching

It is usually not the student who is failing, because that student has already been noticed. It is the one whose score drifts downward across three terms while remaining comfortably acceptable, and the one who arrives at correct answers but takes twice as long as the class to get there. Effort and time are earlier indicators than marks, and both are visible in the way students work rather than in what they finally score, which is precisely what performance analytics should be built to surface.

Catching a gap in the term it appears rather than three years later is not an ambitious goal, and it does not require predicting anything. It requires measuring the right unit — the concept rather than the chapter — often enough that the measurement can still change what happens next. Most of what stands in the way is the shape of the assessment, not the difficulty of the problem. If you are looking at this in your own school, the early-warning side is the natural place to start.