Types of Learning Styles: What Research Says for Training

Go from storyboard to an interactive course in minutes

Start Free

Imagine an operations manager forwarding a new-starter pack with the request, "Can the new course work for every learning style?" It opens with a quiz that sends visual learners to a video, auditory learners to a talk-through and hands-on learners straight to the job. The intention is to personalise training. The question is whether the label provides a sound reason for choosing who gets each activity.

I'm the CEO of Coursebox, and before that I was a learning designer at the University of Adelaide, where I led the design of three edX courses. Every course I design comes down to three kinds of page: a lesson, a practice activity and an assessment. In this guide I set out the common types of learning styles and where each list comes from, what the research says about teaching to them, and what to do instead in workplace training: choose the method by the content, not by the learner. Then I redesign that new-starter course, with a planner you can download and a test of our own product. Coursebox makes an AI tool for employee training that builds one course, with video, text and quizzes, for a whole team, so weigh what I say about it with that in mind.

What are the types of learning styles?

Learning styles are models that sort people into types by how they prefer to take in information, such as seeing, hearing, reading or doing, usually with the claim that teaching matched to each type helps people learn more.

The phrase "types of learning" usually means one of a handful of these models, and the field has no agreed list. A team led by Frank Coffield, reviewing the field for post-school education and training in England, wrote that it had "identified 71 models of learning styles" and "categorised 13 of these as major models". The models below are the ones most often quoted, each with the people who created it and what those authors say about using it.

A few notes on each model, and on what its authors say about it:

• VARK is the model behind most "visual, auditory and kinaesthetic learner" talk. Its questionnaire, its four modes and the evidence on the VARK model need a page of their own, and its Visual mode turns out to be narrower than most people assume.
• Kolb's styles come from the four stages of his experiential learning cycle. The current guide to the inventory says the styles "are not fixed traits but dynamic states" that shift with the situation.
• Honey and Mumford built the most workplace-rooted model. Coffield's team reports that they wrote their own questionnaire because Kolb's inventory "had low face validity with managers", and renamed the four styles.
• Felder and Silverman wrote their model for engineering teaching. In a 2010 essay, Felder said "The point is not to match teaching style to learning style but rather to achieve balance".
• Gardner's intelligences are not learning styles, by Gardner's own account. In a piece reposted by the Harvard Graduate School of Education, he objected to people who "collapse 'multiple intelligences' with 'learning styles'".

Coffield's team judged the 13 major models on reliability and validity, "the minimum standards for any instrument which is to be used to redesign pedagogy". For some of the best-known instruments, the review recommended that "their use in research and in practice should be discontinued".

What are the 3, 4, 7 and 8 types of learning styles?

The numbers come from different lists, not from one theory:

• The four types usually means VARK's visual, aural, read/write and kinesthetic. Kolb and Honey and Mumford also have four styles each, but different ones.
• The three types in many school sources is VAK: visual, auditory and kinaesthetic, which is VARK without read/write.
• The seven types lists usually keep visual, aural and physical, then add verbal, logical, social and solitary.
• The eight types lists add a nature or naturalist type.

The labels that turn four into seven or eight are borrowed from Gardner's intelligences, which Gardner says are not learning styles. A study of 123 education professionals found that most described learning styles through VAK or VARK and through Gardner's intelligences, and that each person defined and used the term in their own way. These popular eight-type lists should not be treated as extra VARK categories or as a single established taxonomy.

Are learning styles real? What the evidence says

Learning styles are real in one sense and not in another. People do have preferences about how information is presented to them, and those preferences are stable enough to measure. What has not held up is the claim that teaching each person in their preferred way helps them learn more.

The clearest statement of the test comes from a review for the Association for Psychological Science by Harold Pashler, Mark McDaniel, Doug Rohrer and Robert Bjork, published in 2008. They called the matching claim "the meshing hypothesis", and they agreed that "the existence of preferences with some coherence and stability is not in dispute". Then they set out what a fair test needs:

1. Sort learners by their style first.
2. Randomly give people from each group different teaching methods.
3. Give everyone the same final test.
4. Look for a crossover: each group does best with the method that matches its style.

They found "virtually no evidence" of that crossover, and noted that several of the studies that did use a fair design "flatly contradict" the matching idea. Their conclusion was that there was "no adequate evidence base" for building learning-styles assessments into general education. They were also careful to add that "many have simply not been tested at all", so the review does not prove every possible style model wrong.

The same test has since been run in professional settings. At a US hospital, 123 internal medicine residents worked through four online modules, each built in two versions: one that started with information and one that started with a patient problem. Their style on Felder and Silverman's sensing and intuitive scale was measured first. According to the published abstract of the 2009 trial, knowledge scores were 83.0 and 82.3 for the two versions, and there was no interaction between style and version (p = .59). The hospital built every module twice, and the styles did not change who learned more from which.

Studies with adults and with VARK point the same way:

• Adults and audiobooks: College-educated adults in a 2015 study by Beth Rogowsky and colleagues were randomly given the same non-fiction text as an audiobook or as e-text. The study found "no statistically significant relationship" between their preference for listening or reading and the format, on a test straight away or two weeks later.
• VARK and grades: Among 426 undergraduate anatomy students, Polly Husmann and Valerie O'Loughlin found that grades were "not correlated with their score in any VARK categories". Some study strategies, such as using the virtual microscope, went with higher grades whatever a student's VARK result.

The fairest summary of the matching studies is a 2024 meta-analysis by Virginia Clinton-Lisell and Christine Litzinger. Across 21 studies with 1,712 participants, it found a small average benefit for matched teaching. But most of the studies that showed the crossover did not meet What Works Clearinghouse quality standards, and the authors judged the effects "too small and too infrequent to warrant widespread adoption."

In other words, about three in four outcome measures did not show the crossover, and most of the studies that did show it fell short of the quality standards. The Education Endowment Foundation's toolkit entry on learning styles, last reviewed in October 2025, rates the approach "Unclear impact for very low cost based on insufficient evidence", because "No studies were identified that met the pre-specified inclusion criteria". That rating is for schools in England, and most of the research above is with students. The one fair test with working professionals in this section, the hospital trial, points the same way.

Why the learning-styles idea persists

If the evidence is this thin, why does the idea survive in so many onboarding packs? Four reasons stand out to me.

• Preference feels like evidence. People notice that a video or a diagram suits them, and it feels like proof. Pashler's team suggested that educators are drawn to the idea partly because they "(correctly)" notice how often one learner gets something from an approach that did nothing for another. What they are often seeing is a format that suits the content, not the person.
• The questionnaires are sold. Some style questionnaires are commercial products, and the publisher of one widely used workplace questionnaire says its results can help predict learning difficulties. Coffield's review found "No empirical evidence of pedagogical impact" quoted in that questionnaire's guide, and no other such studies.
• The research people read mostly endorses it. A 2021 analysis by Philip Newton and colleagues, written for medical educators, found that most recent research papers on learning styles assumed they were a useful approach.
• It shows up as a feature. In workplace training it arrives as a styles quiz in onboarding, persona slides that describe staff as "visual" or "hands-on", and software sold as personalised to each learning style.

The review found frequent endorsement within its defined sample of recent papers. That helps explain why a practitioner can encounter confident recommendations, but it is not a census of everything a professional might read or evidence that matching works.

The request that reaches our own support team is different from the one in the new-starter pack. In the selected support threads from the past year, a handful of course builders from course builders have said learners find a course too heavy on text and want more visuals and interaction. In that selected set, one thread used the words "learning styles": a request to check whether a course suited "neurodiverse learning styles", which is an access question. The complaint is that the course is a wall of text, and the fix is not three versions for three kinds of learner. It is choosing, piece by piece, what each part of the content needs to be shown and practised.

What works instead: match the method to the content

Choose the method from the task and from what learners already know and can access. A sidewall marking needs to be seen and interpreted; a physical procedure needs an appropriate practical attempt. Pashler and colleagues note that instructional methods can vary across disciplines. That does not imply everyone needs identical support, pacing or an identical presentation.

The Education Endowment Foundation makes a similar point: "there is some evidence that cognitive preference and task type may be connected (for example, visualisation is particularly valuable for some areas of mathematics)". The task decides the format. If you have not yet worked out what the job actually needs people to do, a training needs analysis questionnaire is the place to start, before choosing any format.

The example: a new-starter course for tyre fitters

Here is how I would redesign the request at the start of this guide. The example is hypothetical; the business, the people and the numbers are illustrative.

A tyre and wheel-alignment chain has 12 workshops in Greater Sydney, the Illawarra, the Central Coast and the Hunter: Penrith, Parramatta, Blacktown, Castle Hill, Hornsby, Liverpool, Campbelltown, Wollongong, Gosford, Tuggerah, Maitland and Cessnock. About 6 new tyre fitters start each month. Some have worked in other workshops, some are in their first job, and several speak English as an additional language. All can use basic hand tools. What is new to every one of them is the chain's own procedure.

The problem is on the job cards. An internal check found wheel refits with no torque figure recorded, and workshop managers say new fitters finish nuts with the impact gun instead of the torque wrench. Tomasz, the operations manager, wants a short online course that every new fitter does before the first shift, followed by practice and sign-off in the workshop.

Lisa, the people and culture manager, built the current new-starter pack, and it is the one from the opening: a 20-question "What's your learning style?" quiz, then a video for "visual learners", a talk-through with a senior fitter for "auditory learners", and the bay for "hands-on learners". Tomasz forwards it and asks whether the new course can work for every learning style.

By the end of week 2, each new fitter should be able to:

1. State the 5 hoist rules before using a hoist.
2. Read a tyre sidewall and say whether the tyre matches the job card.
3. Refit a wheel using the 6 steps, with the torque wrench, and record the figure.
4. Tell a customer a tyre is at or below the business's 3 mm replace-at mark, and handle "just do the cheap ones" without arguing.

Pricing, warranty, anything legal, suspension faults and wheel alignment go to the workshop manager. This job shows the learning-styles question well because it holds four different kinds of content, and each needs a different method for every fitter. No quiz result changes the fact that torquing a wheel has to be practised on a wheel.

The styles plan and the content plan

The styles plan also costs more to keep current. When the refit procedure changes, someone has to update three versions of it and hope they stay in step.

Four kinds of content, four methods

The four kinds below are my own working categories for this job, not a published taxonomy. Each one gets a lesson, a practice activity and an assessment, and the content decides what each looks like.

1. The 5 hoist rules (rules to recall)

The chain's 5 hoist rules are:

These are fictional lesson inputs, not a complete hoist or wheel-service procedure. The workshop's authorised person checks the actual equipment and vehicle instructions, trains and observes the fitter, and decides when use is permitted. Recalling the five statements or completing three refits is not an independent authorisation to work.

1. Lift only at the vehicle maker's marked lift points.
2. Lock the hoist before anyone goes under the vehicle.
3. Never work under a vehicle held only by a jack: put the stands in first.
4. Safety glasses on whenever the impact gun is in use.
5. Clear the bay and call "Lifting" before raising.

Rules are content people need to bring back from memory at the moment they act, so the practice is recall, not rereading.

• Lesson: One short page with the 5 rules and a photo of a vehicle on a hoist with the lift points marked.
• Practice: Five questions with the rules off screen, such as "A vehicle is up on a jack and you need to look underneath. What do you do first?", with the answer and the reason after each. The same rules come back as three recall questions in the week 1 toolbox talk.
• Assessment: The workshop manager asks for the 5 rules on day 1, before the fitter uses a hoist.

Kai, a new fitter starting at Penrith, does the course on his phone during paid induction before workshop practice. He misses rule 3 in the closed-page check, gets the answer and the reason, and Rhys, the workshop manager, asks him again on day 1.

2. Reading a sidewall (symbols to interpret)

A fitter checks three things on the sidewall against the job card: the size, the load and speed marking, and the week and year the tyre was made. Here the picture is the content. A paragraph describing where the date code sits is harder to use than a diagram with the label next to it, for every fitter.

• Lesson: A labelled sidewall diagram, with each label beside the part it names.
• Practice: Six sidewall photos, each with a job card: does this tyre match the job, and why?
• Assessment: A real job card checked with a senior fitter in week 1.

3. The 6-step wheel refit (a physical procedure)

The chain's refit procedure has 6 steps, with the vehicle still on the hoist for the first two:

1. Clean the hub face and wheel mating surface.
2. Start every nut by hand, then snug them in a star pattern.
3. Position and support the vehicle as the approved vehicle and wheel-service procedure requires for final tightening.
4. Tighten in a star pattern with the torque wrench set to the figure on the job card. The impact gun is never used for final tightening.
5. Write the torque figure on the job card and initial it.
6. Give the customer the manufacturer-approved recheck instruction where required and record it on the job card.

Nobody learns to torque a wheel from a screen, whatever a quiz says about them. The online part can show the procedure and explain why each step is there. The practice has to happen with a wheel.

• Lesson: A demonstration video with narration and captions that shows the 6 steps on a real wheel, with the 6-step card as text on the same page.
• Practice: On the training wheel stand in the workshop on day 1, with a senior fitter. The course says so plainly and does not test the procedure with multiple choice. It ends with one open question: "Why is the impact gun never used for final tightening?"
• Assessment: The workshop manager observes and signs off 3 real refits within 2 weeks.

A trial with new doctors in Denmark tested three ways to prepare for a hands-on procedure, and it also tested whether feeling ready matched being ready.

Case study: Preparing new doctors in Denmark for a spinal procedure

First-year doctors from ten hospitals in eastern Denmark, none of whom had performed a lumbar puncture, needed to prepare for one. A lumbar puncture is a hands-on procedure new doctors are expected to do early, and preparation is usually reading or a video.

• What they did: Each doctor had 15 minutes of preparation in one of three formats, then performed one lumbar puncture on a simulated ward, with an actor playing the patient and a training model for the needle. The formats were a 7-minute video built around the goal of each step, positioning and talking to the patient; a 6-minute traditional expert video of the technical steps; and six pocket-sized pages of written steps with three illustrations.
• Evidence and result: In Henriksen and colleagues' 2018 randomised trial (110 doctors analysed: 31, 44 and 35), two blinded experts rated each procedure from video. Mean scores were 42.8 for the goal-focused video, 40.6 for the traditional video and 38.1 for the written steps (p = 0.01), and the goal-focused group was 2.6 times as likely to pass as the text group. Self-confidence did not correlate with performance in any group.
• Limitation: One attempt on a simulated patient, with no retention test. The two videos did not differ significantly on the main score, the winning video combined several design features, and the plain expert video did no better than text on passing.
• Lesson: For a procedure, a demonstration built around what each step is for helped the whole group, and nobody was sorted by style. Confidence was no guide to skill, which is why the refit is assessed by watching real refits, not by asking fitters if they feel ready.

4. The tread conversation (judgement and conversation)

The chain's replace-at mark is 3 mm of tread, its own rule. Fitters meet three situations again and again:

• A customer in a hurry who says "they'll do another month".
• A customer who wants "the cheapest ones you've got" for a car that tows a trailer.
• A customer who says the last workshop told them the tyres were fine.

A good response shows the tread gauge reading, gives the recommendation and the reason in one sentence, offers the options on the job card, and gets the workshop manager for anything about price, warranty or law. The fitter never argues. This is judgement under a little pressure, so the practice is making choices and hearing the consequences.

• Lesson: A short model conversation, as text and as audio of the same words. Both are there for access, for a noisy workshop or a fitter whose first language is not English, not as a version for "auditory learners".
• Practice: A branching scenario with the three customers, with feedback on each choice.
• Assessment: A role-play with the workshop manager in week 1, using one of the three customers.

The course ends with a mixed check: six questions across the hoist rules, the sidewall and the refit steps, in a new order and with the earlier pages off screen, with the answer and the reason after each.

The whole programme fits in two weeks:

The planner has this course filled in, with blank pages for your own and a checklist for auditing an existing course for learning-styles assumptions.

Practice that helps every learner

Once the method fits the content, a few techniques help people remember it whatever their preferences. A 2013 review of ten study techniques by John Dunlosky and colleagues rated two as high utility: "Practice testing and distributed practice received high utility assessments because they benefit learners of different ages and abilities". Rereading and highlighting were rated low. Most of that research is with students, but the most direct workplace test I know of points the same way.

Case study: Quiz questions inside an Accenture training video

Accenture employees, mostly early in their careers, watch short videos as part of their training, and the training team wanted to know what helps people remember them. The study used an eight-minute introductory video on development and operations (DevOps) that the company used in its own training.

• What they did: 153 employee volunteers were randomly split into four groups: watch the video as usual; discuss it with peers afterwards; discuss it with an instructor's three cue questions; or answer a multiple-choice question about every minute during the video, with the correct answer shown straight after.
• Evidence and result: In Okano, Kaczmarzyk and Gabrieli's 2018 study, 99 of 153 employee volunteers took the next-day test. Questions embedded in the video were associated with about a 14-percentage-point advantage over watching alone; the rounded scores were about 70% and 55%. The guided-discussion group also scored about 70%, and the unguided group about 57%. These rounded figures do not establish that two formats were equivalent. Accenture funded the research and helped recruit volunteers; the authors report no other funder role.
• Limitation: One short video, a knowledge test a day later rather than job performance, and about a third of starters did not return for the test. The questions came with feedback, so this tests recall plus feedback.
• Lesson: In this employee sample, embedded questions with feedback improved next-day knowledge compared with watching alone. The trial did not classify or compare learning styles, and it did not establish an improvement for every person. It is a useful option to test without dividing staff by a style label.

Three more principles carry the rest of the load:

• Spread the recall out. Asking the hoist rules again in the week 1 toolbox talk is spaced repetition: the same recall, days apart.
• Use pictures that carry meaning. Words with a relevant diagram generally teach better than words alone, for learners in general, which is why the sidewall gets a labelled picture rather than a "visual" version. Mayer's principles of multimedia learning set out which pictures help and where the evidence runs out.
• Adjust for what people already know, not for their style. A fitter from another workshop needs less explanation than one in their first job. Under cognitive load theory, worked examples help novices and can get in the way of experts. Building on what people already bring is also a core idea of adult learning theory.

Access options are not learning styles

Some needs are real and are not learning styles. A fitter in a noisy workshop cannot hear a video. A fitter with English as an additional language may want the words in writing. Someone with a hearing impairment needs captions, and a neurodivergent learner may need shorter pages, a clear structure or the chance to check a step again later.

Those needs are met with access options that everyone can use: captions and a transcript on every video, the step card as text beside the demonstration, and a plain, predictable structure. They are not three tracks for three kinds of learner, and nobody needs a label to use them. If someone has an agreed adjustment at work, follow that person's own adjustment.

What to say when someone asks you to cater for learning styles

When a manager asks for training that "caters for every learning style", the most useful answer is about what you will build instead. A 2020 review by Philip Newton and Atharva Salvi found "no evidence that this belief has declined in recent years", and suggested that "It may be better to focus on the promotion of techniques that are demonstrably effective, such as retrieval practice". So I would spend the conversation on the plan rather than on the research.

Here is what I would say, in three points:

• Preferences are real. People can tell you how they like to learn, and that is worth hearing as a conversation about access and comfort.
• Matching has not been shown to help. In fair tests, giving each person their preferred format has not made them learn more.
• So the course is built around the content. Each part gets the method it needs, everyone works towards the same relevant standard, with practice and access support adapted to their needs.

If the request is to sort staff with a questionnaire, the models' own authors are the best reply. The current guide to Kolb's inventory says the instrument "is not intended for use to predict behavior for purposes of selection, placement, job assignment, or selective treatment", and that this includes "not using the instrument to assign learners to different educational treatments". Coffield's team reports that Honey and Mumford's questionnaire "should be used for personal and organisational development and not for assessment or selection".

In the tyre fitter example, that is what happens: Tomasz asks Lisa to take the styles quiz out of the new-starter pack, and the day 1 conversation includes that one question instead.

Building one course for everyone with an AI course tool

On 26 October 2026 I gave the Coursebox course agent the tyre fitter course as a brief. Coursebox is our product, so weigh my account with that in mind. The brief described one course for every fitter, with the four kinds of content and the method each needs: closed-page recall for the hoist rules, a labelled diagram for the sidewall, a demonstration and the wheel stand for the refit, and a branching scenario for the tread conversation. It gave the 5 hoist rules and the 6 refit steps word for word, and told the agent not to sort learners into learning styles or mention them anywhere.

Before it built anything, the agent asked two questions instead of inventing facts. The first was "How should I handle the missing sidewall interpretation content and the six tyre-to-job-card practice items?", since the brief named the sidewall markings but did not explain them. I let it use standard tyre-industry knowledge and invented practice examples for that section only. The second was about the demonstration: "This session cannot create narration, filmed footage, or a video from your procedure." I chose its recommended option: the six-step card and a visual step-through on the page, plus a script for a trainer to film later. So no generic video stood in for a demonstration it could not make.

It then planned a different format for each kind of content. The outline had 7 sections, 13 pages and about 36 minutes, and named each page for what its content needed rather than repeating one template. The hoist rules had a knowledge page, "Five rules before the lift", and a practice page, "Recall the five rules". The sidewall had "Find the three markings" and "Does the tyre match the job?". The refit had "Refit the wheel: six steps", a trainer production script and "One question before practice". The tread conversation had "A calm tread conversation" and "Choose what to say". The outline shows the plan, not what rendered. The agent's hand-over note said the sidewall page uses "a drawn labelled diagram", but apart from the refit page I did not open the pages.

The problem was on the refit page, the one page where the step card is the whole point. Under "Keep these six steps available while watching the demonstration and during supervised practice", every numbered step read the same way: "1. The exact wording for wheel-refit step 1 is not available in the supplied course material.", and so on to step 6, although the brief gave all six steps word for word. The agent's note flagged its own unfinished pages, saying "pages 3, 6, 7, 11 and 12 did not complete correctly". The video area held a placeholder, as agreed: "The demonstration video will appear here with captions and a full transcript".

In the page editor I replaced the six missing steps with the supplied test brief's wording and reloaded the learner view. The card then showed the sequence, followed by the plan for supervised practice and observations. This was a content-recovery check, not verification of a wheel-service procedure. The blog's revised training example uses the approved vehicle and equipment instructions for support position and any recheck interval; that revised wording was not retested in the product.

I checked the outline, the agent's questions and hand-over note, and the refit page after the fix. I did not fix the other placeholders the agent reported: on the recall page, the trainer script, two mixed-check items and the resource page. My check of every page for style language did not finish, so all I can say is that the brief told it not to use any. No video was made, and a real demonstration needs the workshop's own footage. None of this shows that fitters will torque wheels better. The wheel stand and the three observed refits are where that gets checked, outside the course.

Sort the content, not the people

The types of learning styles describe something real: people do have preferences about how they take in information. What the evidence does not support is building training around those preferences, and the authors of the best-known questionnaires say their tools are not for sorting people into different training. In workplace training, the better question is what each piece of content needs to be shown, practised and checked, while prior knowledge, access needs and observed performance guide the support each person receives.

Your next step: take one course you run now and list every piece of content in it. Next to each, write how it is shown, where people practise it and how you check it. If a styles quiz or a "visual version" appears anywhere in that list, replace it with the method the content needs and an access option everyone can use.