The fourth-grade slump

Why do children who read fluently at seven begin to fail at reading by twelve?

A child reads aloud at seven with real fluency: smooth, expressive, barely a stumble. Their scores are fine. Four or five years later the same child is failing reading assessments, and nobody can point to a moment when they got worse.

That last part is the puzzle. A skill that was working does not usually stop working while the person keeps practising it. So either something in the child broke silently, or the thing being measured stopped being the thing they were good at. It is worth finding out which, because the two diagnoses lead to completely different responses.

Start by separating two things that the single word reading runs together.

Decoding is turning marks on a page into words: knowing that these letters make that sound, recognising the word, doing it fast enough that it costs no attention. Comprehension is understanding what the words are saying once you have them.

Now ask what a reading test at seven is actually measuring. At that age, decoding is the hard part and the texts are simple — short sentences about familiar things. A child who decodes fluently will understand nearly everything they decode, because the content is within everyday experience. So the test measures decoding, and reports it as reading.

This is where the neighbouring accounts in this collection come in, and I want to be clear about what they already establish so I can build on rather than repeat them. One shows that once decoding stops being the bottleneck, comprehension is limited by topic-specific background knowledge — you understand a passage about cricket if you know about cricket. Another shows that gaps in vocabulary and knowledge widen over years, because the child who knows more acquires more from the same page.

Both are true, and neither on its own explains a slump. Widening gaps produce a gradually diverging line. Knowledge-limited comprehension produces a ceiling. Neither produces a child who was fine and then was not.

So what makes it look sudden? Here is the move: the texts change.

Somewhere around the middle of primary school, the reading a child is given stops being mostly narrative about familiar situations and becomes mostly expository — science, history, geography, written to convey information the reader does not already have. The vocabulary shifts from words children hear in speech to words that appear almost only in print. Sentences get longer and syntactically denser. And crucially, the text stops supplying its own context: a story about a birthday party tells you what is happening, while a passage about erosion assumes you know what a river is doing.

Now the shape of the failure makes sense. The child's decoding did not decline. The demand simply switched from the thing they had to the thing they did not, and it switched over a couple of school years rather than gradually. The slump is a threshold crossing in the material, not a deterioration in the child.

And that explains the cruellest feature of it: the child is still fluent. They can read the passage aloud beautifully and tell you nothing about it. Which is precisely why it goes undetected for years — every visible sign of reading is intact.

The name is misleading in two ways, and both matter. It is not confined to one grade — the shift is gradual and its timing varies by curriculum, language and child. And "slump" implies a fall. What the data more often shows is a failure to keep rising: the child continues improving in absolute terms while the demand rises faster. Framing it as a decline invites looking for something that went wrong, when the honest picture is a demand curve overtaking a capability curve.

The evidence base is weaker than the phrase's fame. The term comes from work in the 1980s and 1990s, some of it on small or unrepresentative samples, and the effect's size and universality are argued over. It is better attested as a pattern in the kind of difficulty children present with than as a reliable quantitative dip in every cohort. I am not going to quote effect sizes; published values vary enormously with the measure, the population and whether the assessment is norm-referenced.

The obvious remedy is the wrong one, and this is the practically important point. A child failing comprehension looks like a child who needs more comprehension instruction — more practice at finding the main idea, more strategy work. Strategies do help a little and are quickly learned; the difficulty is that they operate on content the child does not have. You cannot infer the main idea of a passage about tectonic plates by technique. The better-supported response is content-rich teaching: build the knowledge and vocabulary the later texts assume, in the years before those texts arrive.

The confound worth naming. Because knowledge tracks what a child has been read to about, taken to see, and heard discussed, this pattern correlates strongly with home background — which means an apparent reading problem is often partly a knowledge-access problem in disguise. That does not make it less real, but it does mean an intervention aimed at reading skill alone is aimed at the wrong variable.

The falsification test. If the mechanism is a change in text demand rather than a change in the child, then a child failing a grade-level expository passage should comprehend a passage of equal linguistic difficulty on a topic they happen to know well — and their decoding accuracy and speed should be unchanged from when they were succeeding. If comprehension were uniformly poor across familiar and unfamiliar topics at equal reading difficulty, the account here would be wrong and something general — decoding automaticity, working memory, attention — would have to be doing it.

• Why comprehension strategies are quickly learned and then stop paying, while knowledge keeps paying.
• How expository text withholds the context that narrative supplies for free.
• Why a fluent reader who understands nothing is the hardest case for a teacher to spot.