Drafted with AI assistance, edited by Yahya Hafez.
Loading post…
Monthly notes on comprehension, accessibility, and AI tutoring. No spam, no upsells, unsubscribe anytime.
Ask two people to retell a chapter they read last week and you will get two different chapters. Not two summaries of the same thing, but two different stories, each shaped to fit what its teller already believed about the world. This is not a failure of attention or a sign of a weak reader. It is how comprehension works. Reading is less like recording and more like rebuilding, and the blueprint you rebuild from is called a schema.
In 1932 the Cambridge psychologist Frederic Bartlett published Remembering, a book built around a deceptively simple experiment (Bartlett, 1932 (opens in new tab)). He gave English students a Native American folk tale called "The War of the Ghosts," a story full of canoes, seal hunting, spirits, and a death that arrives in a way the European mind does not expect. Participants read it, then reproduced it from memory after a short delay, then again days, weeks, and in some cases years later.
The retellings drifted, and they drifted in a direction. Seal hunting became fishing. Canoes became boats. Supernatural events were quietly dropped or given tidy natural explanations. The story got shorter, more coherent, and more English with each telling. Bartlett called this process rationalization, and he argued that people do not retrieve a stored copy of what they read. They reconstruct it from an organized body of past experience he named a schema, and the reconstruction pulls the material toward whatever the schema already contains.
The study was informal by modern standards, with small groups and no statistics. For decades its status was uncertain. In 1999, Elizabeth Bergman and Henry Roediger reran it with Bartlett's own procedure, testing recall after 15 minutes, one week, and six months. They found the same pattern: forgetting over time, plus a growing share of distortions and rationalizations that fit the reader's expectations rather than the text (Bergman & Roediger, 1999 (opens in new tab)). Bartlett's central claim held up.
The vague 1932 idea became a precise one in the late 1970s, when cognitive scientists such as David Rumelhart described schemas as the building blocks of cognition. A schema is a structured package of knowledge about a kind of thing, with variables that incoming information fills in. Your restaurant schema has slots for a host, a menu, an order, a bill, and a tip, along with default values for each. When you read "She left without paying," the schema supplies the unstated facts: there was a bill, and leaving was a violation.
Richard Anderson and P. David Pearson's 1984 review applied this directly to reading (Anderson & Pearson, 1984 (opens in new tab)). On their account a schema does at least four jobs while you read. It directs attention to the parts of the text that matter for the structure you expect. It licenses the inferences the author left out. It tells you which sentences are load-bearing and which are decoration, so memory can be selective. And at retrieval, it gives you a frame to reconstruct from, which is exactly where Bartlett's distortions came from: a frame with no slot for a ghost fills that gap with something ordinary.
This is why the baseball experiment that showed weak readers out-recalling strong ones is a schema result, not just a vocabulary result. The knowledgeable readers were not decoding better. They had a model of a half inning with the right slots, and the text slid into it.
The clearest demonstration of what happens without the right schema came from John Bransford and Marcia Johnson, who in 1972 had people listen to a passage that began: "The procedure is actually quite simple. First you arrange things into different groups..." (Bransford & Johnson, "Contextual prerequisites for understanding," Journal of Verbal Learning and Verbal Behavior, 1972). Every sentence was plain English. Without the title "Washing Clothes," people rated the passage as nearly incomprehensible and recalled under three of its eighteen ideas. Given the title before listening, they recalled about six. Given the title afterwards, it did not help at all. The schema has to be active while the text comes in, not bolted on later.
Reading researchers working with second-language learners found that readers rely on two separate schemas at once. Patricia Carrell and Joan Eisterhold's influential 1983 paper distinguished content schemas, your knowledge of the subject matter, from formal schemas, your knowledge of how texts of a given type are organized (Carrell & Eisterhold, 1983 (opens in new tab)). A scientific paper, a legal opinion, a fairy tale, and a quarterly report each follow a shape, and a reader who knows the shape knows where the claim lives and where the evidence lives before reading a word.
Carrell then tested both at once. In a 1987 study (TESOL Quarterly, vol. 21, pp. 461 to 481), two groups of intermediate English learners, one Muslim Arab and one Catholic Hispanic, read passages whose content was drawn from one of the two religious traditions and whose structure was either a familiar narrative order or a scrambled one. Both kinds of familiarity improved comprehension and recall, content more than form, and readers did worst when both were unfamiliar. Culture, in other words, is a schema too. A text that assumes you grew up with its references will keep its meaning from you even when you understand every sentence.
The neuroscience fits. A 2012 review in Trends in Neurosciences by Marlieke van Kesteren and colleagues argues that information congruent with an existing schema is encoded faster and consolidated more efficiently, because the prefrontal cortex can integrate it into established cortical networks without the slow route through the hippocampus (van Kesteren et al., 2012 (opens in new tab)). Reading with a schema is cheaper for the brain, which is part of why it feels easier.
Schemas are not only helpful. Bartlett's readers remembered a story that was not there. If your model of a topic is wrong, the text does not correct it so much as get absorbed by it. A reader who believes evolution is a ladder toward complexity will read a chapter on natural selection and come away with the ladder intact, because every sentence got filed in the slots the ladder provides. This is the same mechanism that makes re-reading so unproductive: a second pass through the same schema produces the same reconstruction, with the same holes.
So the practical goal is not simply to have a schema before reading. It is to have an accurate one, to activate it at the right moment, and to notice when the text is pushing back against it.
Comprehension is never a transfer from page to mind. It is a negotiation between what the author wrote and the models you brought. You cannot opt out of bringing them, but you can choose which ones arrive and how ready they are to change.
