In defense of knowing things by heart

There is a question that returns every September along with the new pencil cases. Andrés, a history teacher for a tenth-grade class, has been hearing it for years during the first week of the school year, almost always with the same disarming smile: Why do we have to learn this, if I can look it up in a second? This school year, the question has returned with a technical upgrade. It no longer says "look up." It says "ask AI." Andrés knows that the question deserves an answer, because it is a good question. What almost no one has told that student is that the answer exists, that it has sixty years of laboratory research behind it, and that it is far more interesting than the smile.

The short version was written by Gregorio Luri when talking about the machines that now write for us: information is one thing and knowledge is another, and "knowledge will always be a personal construct." Information lives outside and is consulted. Knowledge lives inside and works. With information, you answer tomorrow's exam question; with knowledge, you understand the newspaper, spot the fallacy in an advertisement, and enjoy a cathedral. The difference between the two is not a nuance of vocabulary. It is the difference between owning a library and having read.

The Google Argument, 2026 Version

The disarming-smile argument seemed strong when the search engine took a second. Now that a chatbot writes the entire response, it seems definitive. And yet, it relies on a misunderstanding of how thinking works. No one thinks with what is on the internet. You think with what is in your head—with that, you compare, suspect, relate, and ask better questions. The student who does not know when the French Revolution took place can look it up in an instant, true; what they cannot do is find a novel set in 1793 strange, or notice that the video recommended to them by the algorithm mixes centuries with casual disregard. External culture only yields interest to those who possess internal capital. That is why Luri calls AI products "pre-cooked" thoughts: they nourish, but they do not teach you how to cook. And in a world that cooks pre-packaged food better and better, knowing how to cook becomes the luxury—and also the shield.

This should be said without drama, because the matter does not call for panic. Schools have coexisted before with encyclopedias, calculators, and search engines, and properly used AI has a place in the classroom. What it does not have is the capacity to deliver on behalf of the student the only thing the school exists to produce: having learned. That was already addressed in another piece of this notebook, the one on homework. This one deals with the foundation that the other took for granted.

What a St. Louis Laboratory Discovered

In 2006, two psychologists from Washington University in St. Louis, Henry Roediger and Jeffrey Karpicke, published a series of experiments with an almost domestic design. Groups of students read a text; some reread it afterward, while others were tested on it without looking at the answer. After five minutes, those who reread won: 81% recall compared to 75%. A week later, the scoreboard had flipped—42% for those who reread, 56% for those who were tested. The effort of retrieving something from one's own memory, despite the discomfort it produces, proved to be the action that consolidates it the most. The researchers called it the testing effect, and subsequent literature has confirmed it from all sides.

The finding has a cruel side that any tutor will recognize. In those same experiments, students predicted they would do better by rereading. The strategy that produces the most pleasant feeling of mastery—the text sounds familiar, the highlighting is reassuring—is the one that leaves the worst retention. In 2013, a monumental review commissioned by the Association for Psychological Science to John Dunlosky and his team graded ten study techniques: retrieval practice and spaced study came out with high utility; rereading, highlighting, and summarizing—what the vast majority claim to do—with low utility. To put it clearly: the problem for many students is not that they do not study. It is that they study using the two techniques that work the worst, because no one taught them the others.

For the teacher, this reading is liberating. Testing the lesson—the old practice that modern pedagogy looked down upon—was retrieval practice without knowing it. Properly used, exams do not measure learning. They manufacture it. Luri has been saying this for years, against the caricature that paints him as nostalgic for the podium: the exam as training and not as punishment, the effort of retrieval as the gymnastics of knowledge. The science of memory has been proving him right with an almost comical punctuality.

Memorizing Is Not What They Did to You

A concession is needed here, because the word carries a past. If reading "memorize" made you recall a list of Visigothic kings recited in fear, the critic inside you is right. Rote memorization—without meaning, without understanding, with the ruler threatening in the background—fabricated parrots, and the OECD measured this in PISA 2012: those who said they studied mathematics by memorizing, instead of connecting ideas or planning their study, did just as well as others on easy problems and tanked on difficult ones. But just because copying a hundred times is sterile, it does not follow that memory is an obstacle. It follows that it was poorly cultivated. Retrieval practice is almost its opposite: it extracts outward what has been understood rather than hammering inward what is insignificant, and where copying punished errors, retrieval leverages them. Understanding and memorizing do not compete—they need each other, because no one understands new things except from what they already know. The false war between the two has left two generations convinced that understanding exempts one from retaining, and the evidence says that those who do not retain eventually fail to understand as well.

Memory Furnishes the Imagination

There remains the part of the argument that experiments do not measure and that a humanist faculty should not keep silent about. The verses known by heart, the dates that organize time, the capitals, the formulas, the rivers: all of that is not a warehouse of dead data but the furniture of the inner life. With it, we think when there is no reception and we endure when life gets tough. Nietzsche asked us to read "as cows chew the cud"; Luri adds that true knowledge simmers slowly. What is chewed over stays, and what stays constitutes us—"what we do builds us," he says. Anyone who has learned a language knows this from experience: until the words live in the memory, no conversation is possible, no matter how many dictionaries one carries in their pocket. With general culture, it is exactly the same, except now the pocket speaks. A student who finishes school with nothing in their memory leaves furnished by whichever algorithm is current. One who took inside a handful of verses, maps, and stories leaves with a portable homeland. The difference is noticed at forty, in any conversation that lasts more than ten minutes.

A Faculty That Knows Things

The defense of memory begins, like almost everything in education, with the adult who embodies it. The teacher who recites Machado without opening the book is teaching something that no syllabus includes: that culture can be lived from within. The commitment of Nomba Education goes that way—giving the faculty a deep acquaintance with the humanities, the philosophy that pondered memory, and the literature that deserves to inhabit it, so that the defense of knowledge does not sound like nostalgia but like experience. Meanwhile, Andrés already has his answer for the first week of the school year. AI lends you what it knows, he can tell the one with the smile; yours is what is left when the screen turns off. And then, if the silence lingers, test the lesson. Out of science, and out of affection.

Cover image generated with artificial intelligence.