Learning Styles Myth: What Works Instead

What the Learning Styles Theory Actually Claims

No, learning styles don't work the way they're taught. Sorting yourself into a visual, auditory, or kinesthetic type and then studying mainly in that format has never produced a reliable benefit in a well designed test. The largest synthesis to date puts the benefit at d = 0.04 across 143 studies, which is statistically indistinguishable from zero.

It's worth being precise about what has actually been debunked, because the myth survives partly on a bait and switch.

Nobody disputes that people have preferences. Ask a room of students whether they'd rather watch a lecture or read a transcript and you'll get a real split. Those answers are reasonably stable, too. That isn't the claim under attack.

The claim under attack is what Harold Pashler, Mark McDaniel, Doug Rohrer, and Robert Bjork named the meshing hypothesis in their 2008 review for Psychological Science in the Public Interest: "the claim that presentation should mesh with the learner's own proclivities." In plain terms, that a visual learner will actually remember more if you show them a diagram, and an auditory learner will actually remember more if you say the same thing out loud.

That's the version schools buy. It's the version behind VARK (Neil Fleming's questionnaire, launched in the late 1980s and first written up with Colleen Mills in 1992), Kolb's Learning Style Inventory, Dunn and Dunn's model, Gregorc's Mind Styles, and Honey and Mumford's typology. And it's a big field. Frank Coffield's 2004 systematic review for the UK's Learning and Skills Research Centre identified 71 distinct learning style models, of which 58 were set aside without critical analysis, many of them minor adaptations of a leading model.

Seventy-one models is not a sign of a healthy science. It's a sign of a market.

---

The One Experiment That Could Prove Learning Styles Work

Pashler's team did something more useful than complaining. They specified exactly what evidence would settle the question, and the design is worth understanding because it's fair to the theory.

You need three steps:

1. Classify participants by learning style using whatever inventory you like.
2. Randomly assign people from each style group to each instructional method, so visual learners get split between the diagram condition and the audio condition.
3. Test everyone on the same material with the same test, so the scores are comparable.

Then you look for what statisticians call a crossover interaction. In Pashler's words, the hypothesis "receives support if and only if an experiment reveals what is commonly known as a crossover interaction between learning style and method." The method that works best for Group A has to be a different method than the one that works best for Group B.

Here's the part people miss: this design is generous to the theory. As the authors point out, a crossover interaction "can be obtained even if every subject within one learning-style group outscores every subject within the other." One group can be smarter, faster, better read, and the theory can still win, as long as each group does relatively better with its own format. The test isn't rigged.

So how much evidence met that bar in 2008? Pashler's team scoured the literature and reported that "we found only one study that could be described as even potentially meeting the criteria," a 1999 paper by Sternberg and colleagues, and "even that study provided less than compelling evidence."

Their verdict was blunt: "there is no adequate evidence base to justify incorporating learning-styles assessments into general educational practice. Thus, limited education resources would better be devoted to adopting other educational practices that have a strong evidence base."

---

What Happened When Researchers Ran the Test

The 2008 review was a challenge as much as a conclusion. Several teams took it up and ran the experiment Pashler specified. The results are remarkably consistent.

Rogowsky, Calhoun, and Tallal (2015), in the Journal of Educational Psychology, recruited 121 college-educated adults aged 25 to 40 in the New York City area, all with exactly a bachelor's degree, mean age 30.6. They measured auditory and visual preference with a standardized adult inventory, then randomly assigned participants to learn the same nonfiction book content as either a digital audiobook or an e-text. They tested comprehension immediately and again after two weeks.

Nothing. No statistically significant relationship between learning style preference and instructional method, on either the immediate or the delayed test. The delayed test matters, because it rules out the excuse that the benefit shows up only in durable memory.

Rogowsky and colleagues repeated it with children in 2020, publishing in Frontiers in Psychology. This time it was 118 fifth graders from a rural Pennsylvania middle school, assessed with an inventory built for ages 10 to 13, learning through listening or reading. Their conclusion: "matching instruction to meet a student's auditory or visual learning style had no effect on student achievement."

One detail from that study is quietly damning. Among the 107 children with both comprehension and learning-style scores, the correlation between auditory preference and actual listening comprehension was negative (r = -0.22, p = 0.02). Preferring to listen predicted listening slightly worse, not better.

Husmann and O'Loughlin (2019) ran the most practical version of the test in Anatomical Sciences Education. They gave 426 undergraduate anatomy students the VARK questionnaire, told them their result, encouraged them to adopt study practices matching their dominant style, then surveyed what they actually did and checked it against their grades.

Two findings came out. First, 67% of students failed to study in a way consistent with their own supposed style, despite being handed the recommendation. Second, the minority who did follow it earned no better grades than those who didn't. What predicted grades was specific study behavior: using lecture notes, practising with the microscope. Strategies, in other words, instead of types. The paper's title says it plainly: "Another Nail in the Coffin for Learning Styles?"

---

The 2025 Review That Explains the Confusion

If the experiments keep coming back null, why does the research literature keep producing supportive-looking numbers? John Hattie and Timothy O'Leary answered that in Educational Psychology Review in 2025, and it's the most useful paper on this topic in a decade.

They pulled the 17 meta-analyses on learning styles in the Visible Learning database, together covering almost 700 studies and more than 100,000 students. The average effect size across all 17 is d = 0.40. As they note, that is "precisely the average effect size from 400 influences" in the entire database. On the surface, learning styles look exactly as effective as the average thing you could do in a classroom.

Then they split the pile in two, and the picture inverts.

The four meta-analyses that actually test whether matching instruction to style improves achievement cover 143 studies and 401 effects, and land on d = 0.04. That's zero with a rounding error, and the standard error of 0.21 is five times the size of the effect.

The other 13 are correlational, and their average correlation of r = 0.24 converts to a hefty d = 0.50. They ask whether people who score a certain way on a style inventory tend to do better. That is a completely different question, and Hattie and O'Leary note that this literature "often conflates learning styles with learning strategies." Someone classified as a "deep learner" or a "reflective learner" isn't reporting a sensory preference at all. They're reporting a strategy: how much they elaborate, self-question, and revisit. Strategies do predict achievement. Average the two literatures together, and the strategy effects carry the styles effects until the headline says matching works.

Hattie and O'Leary's conclusion leaves no room: "There is still no evidence for matching, and as many previous reviewers have noted, we should not perpetuate myths, unsupported claims, and wishful beliefs. The recent surge of interest in promoting the correlates of learning styles is misleading, of little value, and should be resisted."

Their prescription is the whole point of this article: shift "toward teaching students adaptable and effective learning strategies that align more closely with task complexity and learning goals."

---

Does Any Study Support Learning Styles?

Honesty requires flagging the strongest counterargument, because it exists and most debunking articles skip it.

In 2024, Virginia Clinton-Lisell and Christine Litzinger published a meta-analysis in Frontiers in Psychology titled "Is it really a neuromyth?" They gathered 21 studies producing 101 effect sizes across 1,712 participants, restricted to work that actually tested matching, and found a statistically significant positive effect: Hedges' g = 0.31, SE = 0.12, 95% CI [0.05, 0.57], p = 0.02.

That's a real result and it deserves to be quoted accurately. It's also much smaller than it first looks.

The confidence interval nearly touches zero at the bottom end, and the paper's own results section reports a slightly different figure from its abstract (g = 0.32, CI [0.07, 0.57]). Only 26% of the learning outcome measures showed matched instruction benefiting at least two different styles, which is the crossover pattern Pashler said was the whole ballgame. And the analysis rests on 21 studies, against the 143 behind Hattie's d = 0.04.

Their own conclusion is not the one the headline invites. Clinton-Lisell and Litzinger judge "the benefits of matching instruction to learning styles" to be "too small and too infrequent to warrant widespread adoption," and recommend that "teaching with multiple modalities may be preferable to the costly and labor-intensive practice of matching instruction to learning styles."

So the fair summary isn't "science has proven learning styles do nothing." It's that the effect, if any survives, is small, inconsistent, rarely takes the crossover form the theory requires, and is dwarfed by techniques that cost nothing to adopt. Even the most sympathetic meta-analysis in the field tells you to teach in multiple modalities and stop sorting people.

---

Do Learning Style Questionnaires Measure Anything?

There's a prior question the effect size debate skips over. Before you ask whether matching helps, ask whether the inventories measure anything stable enough to match to.

Coffield's team took the 13 models they judged major and scored each instrument against four minimum psychometric standards: internal consistency, test-retest reliability, construct validity, and predictive validity. These are the floor, not a high bar. They're what you'd demand of any test before letting it label a child.

One instrument out of thirteen cleared all four. The two models with the deepest commercial reach (Kolb's LSI, and the Dunn and Dunn inventory) cleared exactly one apiece. Coffield's team concluded that six of the models, "despite in some cases having been revised and refined over 30 years, failed to meet the criteria and so, in our opinion, should not be used as the theoretical justification for changing practice."

VARK itself wasn't among the 13 that Coffield graded, so it escapes that particular scorecard. It doesn't escape the outcome data. Husmann and O'Loughlin's 426 students are the largest direct test of VARK's practical value, and it predicted neither what students did nor what they scored.

Coffield's team also flagged the cost of using these labels anyway. Fixed-trait framing "might lead to labelling and the implicit belief that traits cannot be altered," and the advice "that people should work with their strong preferences and avoid their weak ones" leaves learners with a comfortable profile in place of a growing repertoire.

That last point is the real damage. A student who decides she is a visual learner has been handed a reason to skip the reading.

---

Why 89% of Teachers Still Believe in Learning Styles

The evidence has been public since 2008. Belief has barely moved.

Philip Newton and Atharva Salvi's 2020 systematic review in Frontiers in Education gathered 37 samples from 33 studies published between 2009 and early 2020, covering 15,405 educators across 18 countries. The weighted share who agreed that students learn better when taught in their preferred style was 89.1%, ranging from 58% to 97.6% depending on the sample. And 79.7% said they used, or intended to use, the matching approach in practice.

The generational trend runs the wrong way. Pre-service teachers, the ones currently in training, believed at 95.4%. Qualified teachers believed at 87.8%. The people about to enter classrooms are more convinced than the people already in them.

It isn't confined to educators. Nancekivell, Shah, and Gelman reported in the Journal of Educational Psychology in 2020 that 93.7% of a US sample of degree-holding adults, half of them recruited specifically as educators, endorsed learning styles. Reviewing nine prior studies, they put endorsement across the general public and educators in Western and industrialized countries at 80% to 95%.

Their more interesting finding was about the shape of the belief. The soft version is close to universal: 66% agreed that people are born predisposed to a style, 87% that it shows up in childhood, and 80% that it is instantiated in the brain. Roughly half that sample, and two thirds in their first study, went further and held learning styles essentialistically, as fixed at birth and beyond changing. For those people it isn't a study tip. It's an identity, which is exactly the kind of belief evidence bounces off.

There is one hopeful number in this literature. Newton and Salvi found that when educators were given training that explained the missing evidence, belief dropped from 78.4% to 37.1%. Telling people works. It just hasn't been done at scale.

---

What Your Preferences Are Actually Telling You

Throwing out the meshing hypothesis doesn't make your preferences meaningless. It means you've been reading them as the wrong kind of signal.

Preference is a persistence signal, not a memory signal. If diagrams keep you at the desk for 40 minutes instead of 10, that's a real and useful fact about you. Adherence beats optimality for anything you have to do repeatedly. Just don't mistake "I finish more of these" for "I remember more from these."

Aptitude is real, and it isn't a style. Rogowsky's data showed participants classified as visual learners outperforming auditory ones on both measures, listening and reading. Genuine differences in verbal ability exist. What fails is the interaction, the idea that different people need different delivery. Almost everyone benefits from the same well designed instruction.

The content has a modality, even when you don't. A map should be a map. A birdsong should be heard. A dance step should be performed. Richard Mayer's multimedia research is about fitting the format to the material, and that principle is well supported. Mayer himself tested the learner-matching version and found almost no attribute-by-treatment interactions. Dual coding works for the same reason: pairing words with images gives the brain two routes to one idea, and it helps nearly everyone rather than only the visual ones.

The honest translation of "I'm a visual learner" is roughly this: I engage more readily with visual material, so I should use that to stay in the chair, while making sure the strategy I apply to it is one that actually builds memory.

---

What Works Instead of Learning Styles

Strip out style matching and you don't get a vacuum. You get a well-mapped set of techniques with far better evidence, most of which take less time than filling in a questionnaire.

The reference point is Dunlosky, Rawson, Marsh, Nathan, and Willingham's 2013 review in Psychological Science in the Public Interest, which graded ten common study techniques by how well their benefits hold across different learners, materials, and test formats.

Two things stand out. The two high-utility techniques are both about when and how you engage, never about what format the material arrives in. And they're style-blind: active recall and spaced repetition work whether you got the content from a podcast, a paper, or a lecture.

The highlighting rating deserves a caveat, since it's the technique most people reading this actually use. Dunlosky's team rated highlighting as it's typically practised, which is dragging a marker across most of a page. Selective highlighting that forces a decision about what matters is a different behavior with a different outcome, which we've covered in the science of highlighting.

There's also a category the 2013 review couldn't fully capture: desirable difficulties. Struggle that feels unproductive in the moment often produces the most durable learning, which is precisely why studying in your preferred, comfortable format is such a seductive trap.

---

Rebuilding Your Study System Around Strategies

Here's how to convert all of that into a workflow. Notice that none of these steps requires knowing your type.

1. Type the task, not yourself. Before a study session, ask what the material demands. Memorizing terminology, understanding a mechanism, and comparing competing arguments call for different moves. Hattie and O'Leary's central recommendation is aligning strategy to "task complexity and learning goals," which is a question about the work in front of you.

2. Make every highlight a question. The cheapest upgrade available is to stop marking what looks important and start marking what you want to be asked about later. When you highlight a passage with Glasp's web highlighter, add a note phrased as a question instead of a summary. You've just converted a low-utility technique into a retrieval cue.

3. Space the return. Highlighting is capture, and the learning happens when you come back and try to answer before you look. Your highlights become a personal question bank, and revisiting them across days instead of in one sitting turns them into distributed practice for free.

4. Use both channels, for the right reason. Watch the lecture and read the transcript, not because you're a mixed-modality learner but because two encodings of one idea beat a single encoding. Glasp's YouTube video summaries give you the transcript alongside the video, so you can highlight the passage that matters and time-stamp it instead of rewatching 20 minutes to find it. The same logic applies to books: pulling your Kindle highlights into the same place as your web highlights means one review queue in place of four.

5. Explain it to somebody. Self-explanation is a moderate-utility technique on its own, and it gets stronger when the audience is real. Publishing what you've highlighted to Glasp's community forces the clarity that talking to yourself doesn't. If you'd like to rehearse first, Glasp's AI chat lets you interrogate your own highlights and find out where your understanding is thinner than you thought.

Every one of those steps adapts to something measurable: what you've read, what you marked, and what you still can't recall without looking.

---

Frequently Asked Questions

Are learning styles real?

Preferences are real. Learning styles, in the sense that matching teaching to your preference improves how much you learn, are not supported. The four meta-analyses that directly test the matching hypothesis cover 143 studies and produce an effect size of d = 0.04, indistinguishable from zero. The most sympathetic recent meta-analysis found g = 0.31 and still concluded the benefit was too small and too infrequent to be worth adopting.

Should I study in my preferred learning style?

Study in whatever format keeps you working, then apply a strategy that has real evidence behind it. Preference genuinely predicts what you'll stick with, and adherence matters. The mistake is treating preference as a reason to avoid other formats, or as evidence that you'll remember more from one of them.

Is the VARK questionnaire accurate?

It records self-reported preference, and it does not predict which format will teach you more. In the largest direct test, Husmann and O'Loughlin gave VARK to 426 anatomy students and told them the result. Two-thirds didn't study in line with it anyway, and the ones who did earned no better grades.

Why do teachers still teach learning styles?

Because the belief is nearly universal, intuitively appealing, and rarely challenged during training. Newton and Salvi's review of 15,405 educators found 89.1% endorsement, and pre-service teachers believed at an even higher rate than qualified ones. The encouraging finding is that direct instruction about the evidence cut belief from 78.4% to 37.1%.

What should I do instead of learning styles?

Use practice testing and distributed practice, the only two techniques Dunlosky's review rated high utility. Concretely: close the book and try to recall before rereading, and spread the same total study time across more days. Both work regardless of any style you've been assigned.

Does personalized learning work?

Yes, when it personalizes on the right variable. Adapting to what a learner already knows, which questions they keep missing, and how long ago they last saw the material is well supported. Adapting to whether they call themselves visual or auditory is not.

---

The Personalization That Actually Exists

Learning styles got one thing right: generic, undifferentiated instruction really does waste people. It just picked the wrong variable to fix that with.

What differs usefully between two learners is what they already know and which specific ideas haven't stuck. Both are measurable, and unlike a style, both change as you work.

That's also the case for keeping a durable record of your reading. A highlight library is a running map of what caught your attention and what you decided was worth keeping, which makes it usable as a question bank in a way a personality label never was. Glasp exists to make that record easy to build across the web, YouTube, and your Kindle, and to let you see what other people highlighted in the same source. That last part is the piece no questionnaire can give you: a genuinely different perspective on the same page.

Stop asking what kind of learner you are. Start asking what this particular material demands, and whether you can still recall it tomorrow without looking.

References: Clinton-Lisell, V., & Litzinger, C. (2024). Is it really a neuromyth? A meta-analysis of the learning styles matching hypothesis. Frontiers in Psychology, 15, 1428732. Coffield, F., Moseley, D., Hall, E., & Ecclestone, K. (2004). Learning styles and pedagogy in post-16 learning: a systematic and critical review. London: Learning and Skills Research Centre. Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students' learning with effective learning techniques. Psychological Science in the Public Interest, 14(1), 4-58. Fleming, N. D., & Mills, C. (1992). Not another inventory, rather a catalyst for reflection. To Improve the Academy, 11, 137-155. Hattie, J., & O'Leary, T. (2025). Learning styles, preferences, or strategies? An explanation for the resurgence of styles across many meta-analyses. Educational Psychology Review, 37:31. Husmann, P. R., & O'Loughlin, V. D. (2019). Another nail in the coffin for learning styles? Anatomical Sciences Education, 12(1), 6-19. Massa, L. J., & Mayer, R. E. (2006). Testing the ATI hypothesis. Learning and Individual Differences, 16(4), 321-335. Nancekivell, S. E., Shah, P., & Gelman, S. A. (2020). Maybe they're born with it, or maybe it's experience. Journal of Educational Psychology, 112(2), 221-235. Newton, P. M., & Salvi, A. (2020). How common is belief in the learning styles neuromyth, and does it matter? Frontiers in Education, 5, 602451. Pashler, H., McDaniel, M., Rohrer, D., & Bjork, R. (2008). Learning styles: concepts and evidence. Psychological Science in the Public Interest, 9(3), 105-119. Rogowsky, B. A., Calhoun, B. M., & Tallal, P. (2015). Matching learning style to instructional method: effects on comprehension. Journal of Educational Psychology, 107(1), 64-78. Rogowsky, B. A., Calhoun, B. M., & Tallal, P. (2020). Providing instruction based on students' learning style preferences does not improve learning. Frontiers in Psychology, 11, 164. Sternberg, R. J., Grigorenko, E. L., Ferrari, M., & Clinkenbeard, P. (1999). A triarchic analysis of an aptitude-treatment interaction. European Journal of Psychological Assessment, 15, 1-11.