Learning Pyramid: Is It Real? What Research Shows

What Is the Learning Pyramid?

The learning pyramid is a diagram claiming that people retain about 5% of what they learn from a lecture, 10% from reading, 20% from audiovisual material, 30% from a demonstration, 50% from group discussion, 75% from practice and 90% from teaching others. Is it real? No: nobody has found research behind the percentages, Edgar Dale never drew them, and they can't be true as stated. The chart is usually credited to the NTL Institute in Bethel, Maine, and often mislabeled as Dale's "Cone of Experience."

Here's the version you'll see most often:

A second version circulates just as widely, usually printed on top of Dale's cone: we remember 10% of what we read, 20% of what we hear, 30% of what we see, 50% of what we see and hear, 70% of what we say, and 90% of what we say and do. The two charts share a shape and a moral (sitting still is bad, doing is good). They don't share their categories or most of their numbers, which is the first hint that neither one comes from a dataset.

This isn't harmless trivia. The pyramid gets used to argue that lectures are a waste of time, that reading is the weakest way to learn, and that any group activity beats a clear explanation. Each of those claims is either wrong or far more conditional than a tidy triangle suggests. And the pyramid's real lesson, that doing something with information helps you keep it, deserves better evidence than a chart nobody can trace.

---

Where Did the Learning Pyramid Come From?

Follow the citations back and you find plenty of history but no study.

What NTL said. In an email dated February 2003, which James Lalley and Robert Miller reproduced in a 2007 paper in the journal Education, the NTL Institute said the pyramid had been developed and used at its Bethel campus in the early 1960s, when NTL was part of the National Education Association's Adult Education Division. It said it believed the figures were accurate. It also said it no longer had, and could not find, "the original research that supports the numbers." The same reply pointed to a similar pyramid in Dale's revised textbook, but Lalley and Miller noted that the explanation didn't match the figures. Dale's cone, as you'll see below, contains no figures at all.

NTL didn't say the numbers were invented, only that it couldn't produce the evidence. For a claim that's supposed to be scientific, that leaves nothing to check.

The 1967 training-aids article. The "10% of what we read... 90% of what we say and do" figures are usually traced to a 1967 article by D. G. Treichler, "Are You Missing the Boat in Training Aids?", in the trade magazine Film and Audio-Visual Communication. According to researchers who went back to it, Treichler pointed to unnamed studies rather than to any study a reader could look up.

A much older idea. Kåre Letrud and Sigbjørn Hernes went hunting for the pyramid's ancestors and published what they found in Cogent Education in 2018. Their oldest example without numbers came from 1852, in a British debating magazine called The British Controversialist and Impartial Inquirer. Their oldest example with numbers came from 1906, in a piece by Charles Roads about using illustrations to teach the Bible in Sunday schools. They collected 63 versions published between 1906 and 1940, more than two decades before NTL's pyramid supposedly took shape.

Those are the oldest examples they found, not proof of who invented the idea. What they do show is that the claim circulated for about a century, with its numbers shifting as it went, before anyone drew the familiar triangle.

The debunking literature. Learning researcher Will Thalheimer spent years chasing the retention chart back to its sources. In 2014 he and three education scholars, Deepak Subramony, Michael Molenda and Anthony Betrus, published a detailed history of the chart in Educational Technology: "The Mythical Retention Chart and the Corruption of Dale's Cone of Experience."

---

What Edgar Dale's Cone of Experience Actually Shows

A common mistake about the learning pyramid is believing it's the same thing as Dale's cone. It isn't, and the difference matters.

Edgar Dale was an American educator who studied how films, radio and other media could support teaching. In his 1946 textbook Audio-Visual Methods in Teaching, he introduced the Cone of Experience, a picture that arranged kinds of learning experience from the most concrete at the base to the most abstract at the tip. The original bands, bottom to top, were:

1. Direct, purposeful experiences
2. Contrived experiences
3. Dramatic participation
4. Demonstrations
5. Field trips
6. Exhibits
7. Motion pictures
8. Radio, recordings and still pictures
9. Visual symbols
10. Verbal symbols

The revised editions of 1954 and 1969 renamed dramatic participation as "dramatized experiences" and added television, and the 1969 edition lined the cone up with Jerome Bruner's enactive, iconic and symbolic modes of learning.

Notice what the cone doesn't do. It doesn't say how much anyone remembers from anything. It sorts experiences by how concrete they are, on the idea that learners need enough concrete experience for abstract symbols like words to mean something. Dale also told readers not to over-read his diagram. According to scholars who've quoted his editions, he cautioned in 1946 against treating the bands as "rigid, inflexible divisions" or as a "hierarchy or rank order," and called the cone "merely an aid to understanding this subject." In the 1954 edition he described it as a "visual metaphor of learning experiences."

Somewhere along the way, people took Dale's well-known shape, stripped out his categories, and filled it with someone else's percentages. That hybrid is what Subramony and colleagues call the "corrupted cone," and it's still the version most people meet first.

---

Why the Learning Pyramid's Numbers Can't Be Right

You don't need an archive to get suspicious. Look at the numbers themselves.

• They're too neat. Real memory data come out as 47% or 61%, with error bars. They don't land on a ladder of 5, 10, 20, 30, 50, 75 and 90.
• "Retention" is never defined. Retention of what, measured how, after how long? A percentage means nothing without a test and a delay attached to it.
• The categories don't compare like with like. Lectures, reading and video are ways information reaches you. Discussion, practice and teaching are things you do with it. You can read actively, and you can sit silently through a discussion. The "say and do" version even lists reading (10%) and seeing (30%) separately, which led Thalheimer to note: "How people could read without seeing is an obvious confusion."
• No study has measured the whole ladder. You'd need the same material taught seven different ways to comparable learners, tested the same way. No one has identified such a study.

The delay problem alone sinks the chart. In Henry Roediger and Jeffrey Karpicke's well-known 2006 experiment, students who read a short prose passage twice recalled 81% of its ideas five minutes later, 54% two days later and 42% a week later. A group that read it once and then took a practice recall test did worse at five minutes (75%) and better at a week (56%).

So reading has no single retention rate: it depends on when you ask and on what the reader did afterward. A single percentage per activity can't capture that, which is why research reports compare conditions inside one study instead of quoting universal rates.

The kind of test matters just as much. In 1967, Roger Shepard showed people about 600 pictures and then asked them to pick out the ones they'd seen from pairs; the median score was 98%. In Lionel Standing's 1973 study "Learning 10,000 Pictures," people who had studied 200 pictures could describe only about 52 of them in a written recall test, while a two-choice recognition test suggested they knew about 175. That's roughly 26% or 88% of the same pictures, depending on how you ask.

As cognitive scientist Daniel Willingham has pointed out, any memory score depends on what material is being recalled, the delay between study and test, and how memory is tested. The pyramid specifies none of them.

---

How a Chart Without a Source Spread Around the World

With no evidence behind it, the pyramid still ended up almost everywhere.

Its reach is well documented:

• In academic journals. In a 2016 study in the Journal of Curriculum Studies, Letrud and Hernes found 418 peer-reviewed articles published between 1990 and 2013 that presented a version of the pyramid or the retention chart, against just 15 publications that questioned it.
• In medical education. Ken Masters reviewed the medical-education literature in Medical Teacher in 2013 and found 43 journal articles and conference papers using the pyramid. Only one questioned where it came from, and even that one used it to start a discussion. Masters concluded the pyramid should be avoided.
• In myth-busting books. Pedro De Bruyckere, Paul Kirschner and Casper Hulshof gave it a place in Urban Myths about Learning and Education (2015), alongside learning styles.
• Under other names. In Brazil it's widely called William Glasser's pyramid, after the American psychiatrist known for choice theory. A 2018 Brazilian review of its history, by Silva and Muzardo, rejects that attribution, citing Subramony and colleagues' finding that Glasser never referred to or claimed credit for any version of it.
• In many languages. It's ラーニングピラミッド in Japanese, 학습 피라미드 in Korean, pirámide del aprendizaje in Spanish, pirâmide de aprendizagem in Portuguese, Lernpyramide in German and pyramide de l'apprentissage in French, and it's often credited to NTL. Critics have pushed back locally as well, from a Nanzan University paper in Japan on the "fallacy" of the learning pyramid to a medical-education journal at Mexico's UNAM and a University of Regensburg teaching blog in Germany.

Why does it stick? The direction it points feels right: most of us have learned more from fixing something than from hearing about it. The numbers make that hunch look like science. And it's useful to many people at once: to advocates of active learning, to vendors selling training programs, and to anyone who needs a slide that ends the argument. A myth that confirms what its audience already thinks doesn't need evidence to travel.

If this sounds familiar, it's the same pattern behind the learning styles myth: an intuitive idea, a clean diagram, and a claim that survives because so few people check the source.

---

What Research Says About Each Layer of the Learning Pyramid

If you test the pyramid layer by layer, its broad direction holds up: doing something with material usually beats only receiving it. Its numbers and its layer-by-layer ordering don't.

Two warnings before you read too much into that table. First, most of those figures are effect sizes, which measure how far apart two groups end up, not the share of material anyone retained. Second, they come from different comparisons, with different materials and tests, so you can't stack them into a new pyramid. These numbers can't build a new ranking, and they don't rescue the old one either: video, a layer the pyramid credits with just 20%, produced one of the largest effects here when it was added to existing teaching (0.80).

The lecture layer deserves a fairer hearing too. For a beginner, a clear explanation or a worked example is often the fastest way in, as the worked-example studies in the demonstration row suggest. The useful distinction isn't "explanation versus activity." It's teaching that stops at explanation versus teaching that makes students do something with it.

There's also a twist that explains why lectures stay popular. In a 2019 Harvard study by Louis Deslauriers and colleagues, 149 physics students took matched classes taught both ways. They learned more from the active classes, as measured by tests, but felt they had learned more from the polished lectures. Smooth explanations feel like learning. Effortful ones often are.

---

A Better Model: The ICAP Framework

If you want a pyramid that's backed by research, use the one cognitive scientist Michelene Chi and Ruth Wylie published in Educational Psychologist in 2014. Their ICAP framework doesn't sort learning by channel. It sorts it by what the learner is doing:

• Passive: receiving information without doing anything else with it, like listening or reading straight through.
• Active: physically doing something with the material itself, like highlighting, copying, or pausing and rewinding.
• Constructive: producing something that goes beyond the material, like explaining it in your own words, drawing a concept map, or asking why a claim is true.
• Interactive: two or more people building on each other's ideas, where each contribution changes the next.

ICAP predicts that learning improves as you move up the ladder: interactive beats constructive, which beats active, which beats passive. Unlike the pyramid, it has been tested directly. In a 2013 study of engineering students by Muhsin Menekse, Glenda Stump, Stephen Krause and Chi, average posttest scores were 55% in the passive condition, 62% in the active one, 73% in the constructive one and 82% in the interactive one, starting from pretest scores of about 51% to 53% in every group.

Treat those as what they are: short-term test scores from one study, with no delayed test. Chi's own team later found that teachers trained in ICAP still struggled to design truly constructive and interactive activities. A worksheet labeled "discussion" can be passive if nobody has to think.

The big shift is that the same medium can sit at any level. That's why "reading: 10%" never made sense.

Seen this way, the pyramid's moral survives in a better form. What matters is how much thinking an activity forces, whatever the medium.

---

How to Learn More From Reading and Video

You don't need to stop reading or swear off lectures. You need to move up the ICAP ladder while you do them. The routine below fits articles, PDFs, books and videos.

1. Highlight to select, then write to construct. Highlighting counts as an active step. In a 2022 meta-analysis by Ponce, Mayer and Méndez, students who highlighted their own reading remembered more than students who only read (g = 0.36), though their comprehension didn't reliably improve. Push it further by adding a short note in your own words to the lines you mark. With Glasp's web highlighter, you can highlight any web page or PDF and attach the note to the passage. The science of highlighting covers what to mark and what to skip.
2. Turn videos into something you produce. Lectures and tutorials are only passive if you let them be. YouTube Summary gives you a video's transcript with timestamps, so you can highlight the key moment and come back to it. Write your own three-sentence summary first, then compare it with the AI summary. The gaps are what you didn't understand.
3. Test yourself before you review. Close the page and write down what you remember, then check your highlights. Ask Glasp's AI chat to quiz you on what you've highlighted, and answer before you look. Practice testing is one of the two study techniques Dunlosky and colleagues rated highest, and active recall explains how to build the habit.
4. Plan to explain it, then actually do it. In Kobayashi's 2024 meta-analysis, teaching helped when learners knew in advance they'd teach (g = 0.48) and did nothing when they didn't (g = -0.02). So pick someone to explain it to before you start, whether that's a classmate, a colleague, or the people who read the highlights and notes you publish on your Glasp profile or share with the community. The protégé effect goes deeper on why.
5. Bring your books into the loop. Import your Kindle highlights so the passages you marked in books get the same treatment: a note in your own words, a quiz, an explanation.

None of these steps depends on a magic percentage. Each one moves a passive moment at least one rung up the ICAP ladder.

---

Frequently Asked Questions

Is the learning pyramid real?

No. The percentages have never been traced to any study. The NTL Institute, which is usually credited with the pyramid, said in 2003 that it could not find "the original research that supports the numbers," and researchers who have traced the chart's history found versions of the claim, with shifting numbers, going back to 1906. The general idea that doing something with information beats passively receiving it has support. The specific numbers don't.

Who created the learning pyramid?

Nobody knows for sure. NTL says its version was developed at its Bethel, Maine campus in the early 1960s, but it says it no longer has the research behind it. The widely shared "10% of what we read, 90% of what we say and do" figures appear in a 1967 training-aids article by D. G. Treichler, and numbered precursors go back to at least 1906. Edgar Dale and William Glasser are often credited, but neither one published it.

What is the difference between Dale's Cone of Experience and the learning pyramid?

Dale's Cone of Experience, published in 1946, arranges kinds of experience from concrete (direct, purposeful experience) to abstract (verbal symbols). It contains no percentages, and Dale warned against reading it as a ranking. The learning pyramid ranks teaching methods by supposed retention rates, from 5% to 90%. The two are often merged into a "cone of learning" with numbers, which Dale never published.

Do we really remember 90% of what we teach?

No study shows that. Teaching others does help: meta-analyses of learning by teaching put the benefit at effect sizes from about 0.17 to 0.56. That's a real but modest advantage, and it depends on knowing in advance that you'll teach. It isn't a retention rate of 90% or anything else.

Do we only remember 10% of what we read?

No. How much you remember from reading depends on the material, the test and the delay. When Roediger and Karpicke had students read a passage twice in 2006, those tested five minutes later recalled 81% of its ideas, and those tested a week later recalled 42%. Reading followed by self-testing did better over time. What you do after reading changes how much of it you keep.

Did William Glasser create the learning pyramid?

No. The pyramid is commonly attributed to William Glasser in Brazil, but researchers who traced its history found that Glasser never referred to or claimed credit for any version of it. Numbered versions of the claim were already in print in 1906.

What should teachers use instead of the learning pyramid?

Use frameworks that have been tested, such as Chi and Wylie's ICAP framework, which predicts that interactive engagement beats constructive, which beats active, which beats passive. Pair it with strategies that have strong evidence: retrieval practice, spaced practice, worked examples and explaining ideas in your own words. Dunlosky and colleagues' 2013 review rated practice testing and distributed practice as the two highest-utility study techniques.

Is active learning better than lectures?

On average, yes, but not because lectures teach nothing. Freeman and colleagues' 2014 meta-analysis of 225 studies found active learning raised exam scores by about 0.47 standard deviations and cut failure rates from 33.8% to 21.8%. Clear explanations still matter, especially for beginners. The best classes combine them with tasks that make students retrieve, explain and discuss.

---

Keep the Arrow, Drop the Numbers

The learning pyramid gets one thing roughly right and almost everything else wrong. Its arrow, pointing from passive receiving toward active doing, matches what decades of research suggest. Its numbers have never been traced to a measurement, its source can't be found, and Edgar Dale, whose name is so often attached to it, drew something else entirely.

So keep the arrow and drop the numbers. The next time you read an article, watch a lecture or open a book, ask yourself one question: what am I doing with this? If the answer is "just looking at it," move one step up. Highlight the line that matters, explain it in your own words, test yourself tomorrow, and tell someone about it.

Glasp is built for exactly that loop: highlight what you read and watch, add your own thinking, and share it so learning becomes something you do with other people, not just something that happens to you.