Claims about teaching, and what their sources say
Reference

Learning myths and other claims about teaching, and what their sources say

Seven claims you will hear from someone confident, set against the published source behind each one. This is the page to read before you build a session around something you were told.

Seven claims you will hear from somebody confident, each set against the published source behind it. The point is not that the people repeating them are foolish. Most of these are repeated by people with real teaching experience, and two of them appear on pages published by universities and by training companies. The point is that you can check, and that what the sources say is often narrower and more useful than the claim.

Each row gives the claim as people say it, what the source actually says, how strong the evidence is, where the claim came from, and what to do instead. Filter by how the claim is usually used.

The claims

The claim, as people say itWhat the source saysHow strongWhere it came fromWhat to do instead
"People remember 10 percent of what they read and 90 percent of what they do." The retention percentages attached to reading, hearing, seeing, doing and teaching trace to no research. Versions of the numbers appear as early as 1914 and 1922, the corruption of Edgar Dale's Cone of Experience is dated to around 1970, and Dale's 1946 cone makes no claim of scientific grounding. The cited sources are described as false, and the chart as unsupportable in terms of scientific validity or logical interpretability. No supporting study has been produced A practitioner-researcher's investigation summarizing a 2014 Educational Technology special issue. This is one researcher's own site, not peer review; the weight of the point rests on the absence of a source for the numbers rather than on his authority. Design around people doing the thing because practice with feedback is supported in its own right, not because of a pyramid. See Designing the part where they do it.
"Attention drops after 10 to 15 minutes, so change activity every 10 minutes." A peer-reviewed review traced the claim and reports that the literature is replete with works propagating a 10 to 15 minute attention span; that note taking, the usual proxy, is not a good proxy for attention and would not support such a limit even if it were; and that the claim amounts to an unsubstantiated mantra. The famous 8-second figure is traced to a study commissioned by Microsoft and reported in Time, and actually describes the average time a person spends on a web page. Traced and unsupported. The counterclaim is a university teaching center page that still states the 10 to 15 minute figure, which is how widely it is held Repeated in teaching guidance for decades. One university teaching center's active learning page states it; Bradbury (2016) traced it and found nothing behind it. We show both and name the disagreement. Segment for a reason you can name: a check, a practice block, a change of posture. Not for a clock. See Structuring a session: the run sheet.
"Teach to each person's learning style." A review of the evidence found virtually no support for the interaction pattern the idea requires, with several adequately designed studies flatly contradicting it, and concluded there is no adequate evidence base to justify incorporating learning-styles assessments into general educational practice. The same authors state that it would be an error to conclude that all possible versions of learning styles have been tested and found wanting. A review in a public-interest series. We state its caveat every time we state its conclusion Pashler, McDaniel, Rohrer and Bjork (2008). The idea remains widely taught: one professional association's training certificate publishes two topics about matching delivery to learner preferences, though its page does not say what those topics teach. Vary the format because the material has different parts, not because people are different types. Show it, say it, write it down, and let people practice.
"Active learning beats lecture, so never lecture." Performance on examinations and concept inventories increased by 0.47 standard deviations under active learning across 158 studies, and students in traditional lecture courses were on average 1.5 times more likely to fail across 67 studies. A 0.47 standard deviation change corresponds to about a 6 percent increase in average examination scores. A meta-analysis, and a strong one, of a specific thing Freeman et al. (2014). Every study in it is an undergraduate course in science, engineering or mathematics, compared over a term. Build the session around people doing things, and say what the figure covers whenever you cite it. It compares course designs over months, not individual minutes, and it is not a measurement of a two-hour workshop.
"Demonstrate, narrate, have them talk you through, then let them do it: they will retain it." In the trial that tested those four steps, 95 medical students were randomized. No difference on the multiple-choice exam. The Peyton group scored significantly higher on the practical exam. A second practical exam showed a significant decline in competency regardless of instruction type: skills retention was equally low for both techniques. One randomized trial for the practical-exam result; the retention finding is from the same trial Gradl-Dietsch et al. (2016). A 2025 trial that favors the approach states in its own words that there is limited evidence supporting its superiority over other teaching methods. Use the four steps because they force you to narrate and to hand over the doing, which is what Showing someone how, so they can see it asks for. Plan for retention separately: what leaves the room with people matters.
"They loved it, so it worked." Reactions are strongly connected to post-training motivation and moderately connected to self-efficacy, even controlling for pre-training levels. They are only modestly related to immediate knowledge outcomes, unrelated to delayed knowledge, and the connection with skill development is unknown. Reactions are more positive when the trainer uses communication that creates connection: eye contact, moving around the room, using names, personal examples and humor. A summary of meta-analytic evidence, from a federally funded center QIC-WD umbrella summary (2022), citing Sitzmann et al., Alliger et al. and Blume et al. Keep the form and label what each question measures. Add one observation of performance, which Did the training work? Finding out afterward, and changing the session explains.
"Ninety minutes, twenty minutes, ten minutes: that is how to chunk a session." The 90-20-10 rule appears on a commercial training company's boot camp page as one of seven things participants will discover. The page gives no source for it. One company's convention, published without a source The Bob Pike Group's Train-the-Trainer Boot Camp page. We have the marketing page and not the classroom, so we do not say what that program teaches about it. Use the session budget calculator, which has no default segment length, and segment for a reason you can name.

How to check one of these yourself

Four questions settle most claims of this kind, and they take about ten minutes.

  1. Who published it, and is the page the source or a repetition of one? A teaching center page restating a figure is a repetition. The study it cites is the source, and it may not say what the page says.
  2. What exactly was measured, in whom, over how long? A figure from undergraduate courses compared over a semester is not a figure about your two-hour session, even when it points the same way.
  3. Does the source carry its own caveat? Several of the rows above do, in the authors' own words, and the caveat is usually dropped by whoever repeats the claim.
  4. What would the claim predict, concretely, in your room? If you cannot say, the claim cannot change your plan, and a claim that cannot change your plan is not worth arguing about.

Every source behind this page is listed, with the date it was read, in the resource library.

What this manual does with claims

  • Where two sources disagree, we show both and name the disagreement. The attention-span row is the clearest case: a university teaching center and a peer-reviewed review say different things, and we do not quietly pick a side.
  • Where a source carries a caveat, we repeat the caveat every time we use the source. The learning-styles row is the one where this matters most.
  • Where no source gives a figure, our calculators have no default. The buffer, the segment length, the number of practice attempts and the ratio of people to instructors are all inputs with nothing filled in.
  • We do not say what a training provider teaches. Their published pages say what topics they cover, not what they say about them. Where a published topic sits oddly beside the evidence, we name the tension and stop there.