Learning styles are a myth: what the studies show, and what works instead

You do not have a visual or an auditory learning style. You have preferences, and preferences feel real. But matching instruction to a diagnosed style does not reliably improve learning. That is what the research finds, from a 2008 review to a 2024 meta-analysis.

• Learning styles describe preferences for how information arrives. They do not describe how you learn best.
• The claim that needs proof is the meshing claim: each style learns best in its own format. Almost no study passes the test for it.
• In a study of 125 fifth graders, 68 percent had no clear auditory or visual preference, and matched instruction had no effect on scores.
• A 2024 meta-analysis of 21 studies found a small average matched advantage, but the predicted pattern in only 26 percent of measures, in mostly low-quality studies.
• What helps instead is spaced practice, retrieval, interleaving, and multimodal teaching chosen by content. Our guide to memorising psychology terminology puts the first two to work on psychology terms.

What people mean by learning styles

In English, the best known model is VARK. Neil Fleming and Colleen Mills described it in 1992. It sorts preferences into four modalities: Visual, Aural, Read/write, and Kinesthetic. Visual means maps and diagrams. Aural means discussion and listening. Read/write means lists, notes, and text. Kinesthetic means examples, practice, and hands-on experience.

Most people are not one type. The official VARK site says most people are multimodal and use several modalities as the situation changes. It also says the result shows preferences, not a fixed label:

The same page says VARK does not measure intelligence and does not predict success. It makes no claim that matched teaching improves learning. That restraint matters, because many secondary pages claim more than the model does.

In German schools, the parallel tradition is the Lerntypen of Frederic Vester. His book Denken, Lernen, Vergessen appeared in 1975. As quoted by the didactics researcher Maike Looß, Vester described four types: auditory, visual, haptic, and learning through the intellect. His example is the law of pressure, force through area. Each type reaches it through another channel. Looß notes two problems. The fourth type is not a sense channel at all. And Vester himself recommended many channels for every pupil, not one channel per pupil. The popular matched-only version came later.

The test a learning-styles study has to pass

Harold Pashler and colleagues set the test in a 2008 review in Psychological Science in the Public Interest. The claim has a name: the meshing hypothesis. Instruction works best when its format fits the learner's style. To prove it, a study must do three things.

1. 1.Sort learners into at least two style groups with a style measure.
2. 2.Assign learners from each group at random to at least two instruction methods, such as listening and reading.
3. 3.Give every learner the same final test, then look for a crossover: the method that is best for one style group must not be best for the other group.

The crossover is the point. If one method is best for every group, the study shows that method is better, not that styles matter. Pashler and colleagues found ample evidence that people report preferences, and plentiful evidence that aptitudes differ. For the crossover pattern they found virtually no evidence. Their conclusion was direct: there is at present no adequate evidence base to justify learning-styles assessments in general teaching practice.

What the studies found

The adult study came first. Rogowsky, Calhoun, and Tallal followed the Pashler design with college-educated adults in 2015. They measured auditory and visual preferences with a standard adult inventory. Learners then received the same nonfiction material as a digital audiobook or as e-text, with a written comprehension test at once and after two weeks. Preference did not predict aptitude in part one, and preference did not interact with method on either test in part two.

The classroom study with children is the sharpest test. Rogowsky, Calhoun, and Tallal repeated the design with fifth graders in 2020. Twelve children were auditory, 22 visual, and 73 (68%) had similar preferences across modes. Most of the grouped children therefore had no clear type. Children took listening and reading versions of a standard comprehension test in counterbalanced order. Visual-preference children scored higher than auditory-preference children in both modes. There was no instruction by style interaction. The authors write that matching instruction to auditory or visual preference had no effect on student achievement.

The anatomy study tests VARK directly. Husmann and O'Loughlin surveyed 426 undergraduates in 2019. Students completed the online VARK questionnaire and a study-strategies survey. Most students did not use strategies that fit their VARK result. Course performance had no link with any VARK score. Some single strategies, such as the virtual microscope, linked with higher grades for every student, whatever their VARK. Fit between strategies and VARK had no link with outcomes.

The judgments study explains why the label sticks. Knoll, Otani, Skeel, and Van Horn found in 2017 that verbalizer scores went with higher immediate judgments of learning for words, and visualizer scores with higher judgments for pictures. Neither score predicted recall or the accuracy of the judgments. Style tracked the feeling of learning in the moment, not the result on the test.

The 2024 meta-analysis puts numbers on the whole field. Clinton-Lisell and Litzinger screened thousands of records and kept 21 studies with 101 effects and 1,712 participants, as reported in Frontiers in Psychology. The average matched advantage was g equal to 0.31, with a confidence interval from 0.05 to 0.57. But only 26 percent of outcome measures showed the crossover for at least two styles. Of 12 further studies without enough statistics, 25 percent pointed the same way. Most studies did not meet What Works Clearinghouse standards, and 85 percent of the studies in the analysis dropped participants whose scores allowed no confident type. The authors interpret the benefit as too small and too infrequent for wide use in teaching.

Why the label feels right

Preferences are real. Pashler and colleagues never denied that people report them. VARK exists because people differ in what they like to receive. The error starts when a preference becomes a prescription: teach me only through my channel.

Three facts keep the myth alive. First, tests return an answer. A quiz that asks how you like to receive information will give you a type, and a type feels like a diagnosis. Second, the type predicts confidence. As Knoll and colleagues show, it tells you how sure you will feel now, not how much you will keep later. Third, almost every teacher believes it. Dekker, Lee, Howard-Jones, and Jolles found in 2012 that 93 percent of UK teachers and 96 percent of Dutch teachers agreed that people learn better in their preferred style. Newton and Salvi counted 37 samples with 15,405 educators from 2009 to 2020 and found a weighted belief of 89.1 percent in 2020. Nancekivell, Shah, and Gelman found in 2019 that more than 90 percent of 668 participants believed in styles, with some holding the view as fixed at birth, as the American Psychological Association reports.

Newton and Miah add the use behind the belief. In a 2017 survey of 114 UK academics in Frontiers in Psychology, 58 percent believed in styles but 33 percent had used them in the last year. Ninety percent agreed the theory has a basic conceptual flaw. It cannot teach understanding in one channel, because many ideas, such as heart sounds or skin rashes, live in one modality by nature. About one third said they would keep using styles even after reading the evidence.

Willingham, Hughes, and Dobolyi put the distinction in one line in 2015: styles are preferences, not abilities, and scientific support for the theories is lacking. That sentence is the safest answer when someone asks whether styles exist. Preferences exist. The matching benefit does not.

What to use instead

Stop asking which channel you are. Start asking which method the material needs. The same review literature that rejects matching endorses a short list of methods that help most learners.

1. 1.Space your practice over time. If an exam is in three weeks, study a topic in several short sessions across the weeks instead of one long session the night before. The gap between sessions matters more than the channel. Our article on spaced repetition shows the schedule, and the forgetting curve shows why cramming fades fast.
2. 2.Test yourself by recall. Close the book and write what you remember, then check. Dunlosky and colleagues rated practice testing and distributed practice as high utility in 2013 because they help learners of many ages and abilities across many tasks.
3. 3.Mix related topics in one session. If you study conditioning, mix classical examples, operant examples, and schedules instead of blocking each one alone. Mixing feels harder and scores higher later. Our article on interleaving shows when it helps and when it does not.
4. 4.Use a memory route for ordered lists. If you must keep an order, such as the four VARK types or the three Pashler steps, place one vivid image per item along a route you know. Our article on the method of loci gives the five steps and an exam example.
5. 5.Teach every learner with words and pictures together, chosen by content. Richard Mayer's multimedia work finds that people learn better from words and pictures than from words alone, as reviewed in 2024. A heart sound needs audio. A skin rash needs an image. A formula needs both plus practice. That is multimodal teaching, not matching.

Studying with Psychology World

Save the pairs you confuse into one named bookmark list, such as preference against style, or meshing against multimodal. Then drill that list with Flashcard, Question, and Typing until each answer comes fast. Daily-word pushes bring you back, so the list stays warm until the exam.