VARK Model: Learning Styles, Questionnaire and Evidence

Go from storyboard to an interactive course in minutes

Start Free

An area manager at a café chain comes back from a leadership day where everyone took a learning-styles quiz. Her next email asks the training team to rebuild the closing-clean course for new baristas so that it "works for visual, auditory and kinaesthetic learners", ideally as three versions. The request is well meant. She wants every new starter to be able to close the store properly, and the quiz gave her a way to talk about that.

I'm an instructional designer at Coursebox, where we make instructional design software that turns a brief into a course outline. One section of this guide tests that tool, so read it knowing it is our product. Requests like the area manager's are why I wrote this. I cover what the VARK model is in Neil Fleming's own words, what each letter means, where to take the official questionnaire, what the research says about matching training to preferences, and what I would build for that café instead, with two free templates.

What is the VARK model?

The VARK model is Neil Fleming's description of four preferences for taking in and giving out information (Visual, Aural, Read/write and Kinesthetic), explored through a short questionnaire he launched in 1987 as a prompt for learners to reflect on how they study.

View the full-size diagram

Fleming developed it at Lincoln University in Canterbury, New Zealand, and first wrote it up with Mills in 1992. In that paper in To Improve the Academy, they describe "a modal preferences questionnaire as a catalyst to empower students to reflect on their own sensory preferences and modify their study methods accordingly". So VARK began as a study-skills tool for learners. It was not a system for teachers to sort a class.

The structure is simple. The questionnaire describes everyday situations, and you can choose more than one answer for each. Your answers are added up for each mode, and the result is a profile: one mode that stands out, or a combination of two, three or all four. The modes are not ranked, they are not stages, and they do not exclude each other. Fleming and Mills even note that Kinesthetic "is not a single mode", because practice and real examples can involve all the senses.

Here are the four modes as VARK defines them today.

Most people do not land on a single mode. VARK's publisher now describes two multimodal patterns: Selective, where someone switches between modes depending on the situation, and Integrative, where someone tends to use several modes together. Older sources call these "Type One" and "Type Two".

Who created VARK, and what did Fleming say it was for?

Neil Fleming was a school inspector in New Zealand before he moved into staff development at Lincoln University. In a 2006 article with Baume, he wrote that he "watched some 9000 classes" over nine years as an inspector, and was puzzled when "excellent teachers" did not reach some learners while "poor teachers" did. He launched VARK in 1987. The 1992 paper says the starting point was a set of visual, aural and kinesthetic categories from neuro-linguistic programming literature, and Fleming's addition was to split the visual category in two so that reading and writing had a mode of its own.

The dates to use when citing it:

• 1987: Fleming launches VARK at Lincoln University.
• 1992: Fleming and Mills publish the first paper, with a 13-question version.
• 2012: the name VARK is trademarked, according to the copyright page on VARK's site.
• 16 June 2022: Fleming dies in Christchurch, aged 82, as VARK's own team page records. VARK Learn Limited continues the work.

Fleming was consistent about what VARK was not. In 1992, he and Mills wrote that "it is simply not realistic to expect teachers to provide programs that accommodate the learning style diversity present in their classes", and that the profiles were "promoted as insights and not as definitive diagnoses". They added that teachers could still consider learning styles when planning, but put the weight on learners adjusting their own study. Because the questionnaire was meant to start discussion, they called testing it for validity and reliability "unnecessary and inappropriate".

In 2006 he went further. VARK is "technically, not a learning styles questionnaire", he wrote, because it reports only on preferred modes for communicating, and it is "a catalyst for metacognition, not a diagnostic or a measure". He also warned against reading a preference as a skill. Some people, he wrote, "confuse preferences with ability or strengths", adding: "You can like something, but be good at it or not good at it or any point between." VARK, in his words, "tells you nothing about the quality of that communication."

There is one nuance to keep. Fleming did believe learners should line up their own study strategies with their preferences. In the same 2006 article he wrote that information accessed through strategies "aligned with a student's modality preferences is more likely to be understood and be motivating", before adding, "But I am not a theorist." He offered no data for it. That claim is about a learner choosing how to study, not a trainer sorting a group, and it has since been tested.

VARK's publisher holds the same line today. Its page on what VARK is and isn't carries the heading "VARK Does Not Advocate 'Matching' Teaching to Learning Styles", and its guide for teachers and trainers lists "Trying to match teaching styles to each learner" first among common mistakes.

Which of Fleming's papers to read

• Read the 2006 article by Fleming and Baume first. It is four pages, free on VARK's site, and gives Fleming's own account of what VARK is for and where it stops.
• Read the 1992 paper by Fleming and Mills for the origin. It explains the four modes, the reasoning behind the questionnaire and why the authors did not want it treated as a diagnosis.
• Use VARK's "What VARK is and isn't" page for the current position. It is the publisher's own statement on matching and on what the results do and do not tell you.

The VARK questionnaire: how to take it and what your results mean

The official questionnaire is on VARK's own site. At the time of writing it is version 9.2, with 16 everyday scenarios, and you can tick more than one answer for each. There is also a version for younger people. Taking it for your own study is free.

Your result is a profile, not a type. VARK's introduction says the results reflect "your unique pattern of preferences rather than a fixed label", and that VARK does not measure intelligence or predict success. A single strong mode is unusual:

Those figures come from the research statistics page on VARK's site. The shares of two-mode (20.1%), three-mode (15.0%) and four-mode (31.0%) profiles are shares of all 1,048,292 respondents, not of the multimodal group. The most common single profile was the Integrative four-mode pattern, where all four modes are used together, at 25.4%.

So what can you do with a result? The use Fleming intended is personal: notice how you like to take in information, then try study strategies that use that mode and others. Treat a result as a starting point for that conversation, not as a reason to avoid a format.

Does matching training to VARK preferences improve learning?

Not reliably. The studies that test it fairly have found little or no benefit. The idea that people learn more when material comes in their preferred mode is often called the matching or meshing hypothesis. In a 2008 review in Psychological Science in the Public Interest, Pashler and colleagues set out what a fair test needs. People are sorted by their preference, then randomly given matched and mismatched versions of the same material, and everyone sits the same test. Matching is supported only if each group does best with its own format, a crossover. Across learning-style models, they "found virtually no evidence" of that pattern.

Case study: Matched and mismatched formats in US comprehension trials

These trials tested the listening versus reading version of the matching idea, using a different preference questionnaire from VARK. Rogowsky, Calhoun and Tallal asked whether people understand more when material arrives in the sense they say they prefer.

• What they did: In a 2015 study of college-educated adults, participants were randomly given the same nonfiction text as a digital audiobook or as an e-text, then tested on comprehension straight away and two weeks later. In a 2020 follow-up in a rural Pennsylvania school, every child in the fifth grade sat both a listening and a reading comprehension test, in counterbalanced order.
• Evidence and result: The adult study found "no statistically significant relationship" between auditory or visual-word preference and format, for immediate or delayed comprehension. In the school study, 125 children took part and 107 had usable preference and test data. Of those 107, 73 had similar auditory and visual preferences and could not be put in either group. Among the 34 who could, there was no interaction between preference and format (p = 0.29), and both groups scored higher on listening than on reading.
• Limitation: The preference measures came from the Dunn and Dunn family of inventories, not VARK, and covered listening versus reading words only, with no diagrams or practice. The tasks were short comprehension tests, not job training, and the 2020 sample was children.
• Lesson: When people were sorted by the sense they prefer and given matched and mismatched formats, the match did not help. In the school cohort, most children could not even be sorted.

The main VARK-specific study tested Fleming's claim that learners should study in line with their preferences. Husmann and O'Loughlin surveyed 426 undergraduate anatomy students at Indiana University, who completed the VARK questionnaire and a survey of how they studied. In their 2019 paper, most students did not report study strategies that lined up with their VARK result, and "student performance in anatomy was not correlated with their score in any VARK categories". Some strategies, such as using the virtual microscope, went with higher grades regardless of anyone's VARK profile. The study was correlational, so it shows no link rather than proving one could never exist.

A 2018 mini-review by Aslaksen and Lorås looked specifically at modality-based styles, the kind VARK describes, and found "no replicable statistical evidence" that aligning instruction with them improves learning outcomes.

The fair counterpoint is a 2024 meta-analysis that did find a small average effect:

In that meta-analysis, Clinton-Lisell and Litzinger concluded that the benefits were "too small and too infrequent to warrant widespread adoption". They set it beside the modality effect, where narration explains a picture, which they cite at about twice the size for all learners, with no sorting needed. The most supportive workplace-style study I found is a 2021 simulator study of trainee pilots at an Australian university, which reported that matching feedback to each pilot's VARK preference improved performance. It involved 18 pilots, and the feedback mode was partly tied to which part of the flying it addressed, so it is a reason to run a small trial, not to rebuild a curriculum.

I could not find an adequately sized trial in a workplace that sorted employees by VARK and gave them matched and mismatched training. For an L&D team, that gap is the finding: the case for building style versions of a course has not been made where you would need it.

The Education Endowment Foundation, which reviews evidence for schools in England, rates learning styles as "Unclear impact for very low cost based on insufficient evidence", and warns that labelling learners "is likely to undermine their belief that they can succeed through effort".

Why VARK still feels right

Belief in matching is very common among the people who design teaching. A 2020 review by Newton and Salvi pooled 37 surveys of 15,405 educators in 18 countries and found that 89.1% believed in matching instruction to learning styles. Most of those surveys cited VARK.

Part of the reason is that the experience of a preference is real. In a 2017 study by Knoll and colleagues, which used a verbal versus visual style questionnaire rather than VARK, people's style predicted how well they expected to remember words or pictures, but not how much they actually recalled. The authors summed it up as a link with "subjective aspects of learning but not objective aspects of learning". VARK's own research questions point the same way: of about 93,000 people who answered them, 72% said their result matched how they see themselves.

So a manager who asks for VARK-based training is not being careless. They are repeating something most educators believe, about a preference they can feel. The preference exists. What the research does not support is the step from "I prefer it" to "I learn more from it".

The worked example: a café chain's closing clean

To show what to build instead, I'll carry one hypothetical example through the rest of this guide, the area manager's request. A café chain with about 40 stores trains new baristas in their first two weeks. They have done one supervised shift, they can make drinks, and they have never closed the store. Some are studying English as an additional language. They complete a short course on their phone before their first closing shift, then do the clean with a supervisor watching.

The closing clean of the espresso machine has five parts:

1. Backflush each group head with cleaning powder, using a blind basket.
2. Purge and wipe the steam wand.
3. Soak the portafilters and baskets.
4. Wipe the drip tray and the machine.
5. Record the clean on the closing checklist.

The supervisor signs off a new barista when they complete the clean in the correct order, with no milk left in the steam wand and the checklist filled in. To make that judgement fair, the course gives learners what they need: a labelled diagram of the machine, the written steps, a demonstration, the checklist itself, and the store's rule for what to do when the cleaning powder runs out at close.

The area manager has asked for the course to work for "visual, auditory and kinaesthetic learners", in three versions. Her real goal is that every new barista can do the close. This example is a good test of a VARK request because each part of the job already has a format that follows from the content, not the learner, and because the outcome can be seen on shift.

The four VARK modes, worked through one closing clean

Here is what each VARK mode would look like in the closing-clean course, and why every barista gets all of it.

Visual: a labelled diagram of the machine

VARK's Visual mode is information shown as diagrams and symbols, what Fleming and Mills called "graphical and symbolic ways of representing information". It suits content with a structure or a layout.

• What VARK counts here: A labelled diagram, a flow chart or a layout, not a photo or a video.
• In the closing-clean course: One labelled diagram of the group head, blind basket, portafilter and steam wand, used again on the steps page.
• Why everyone gets it: The machine is a physical layout. Every barista needs to know which part is which before they can follow a step that names it.

Aural: the reason, explained in a voice

VARK's Aural mode is information that is heard or spoken, including talking something through. It suits explanations and reasons.

• What VARK counts here: Narration, a conversation or a spoken explanation.
• In the closing-clean course: A short narrated explanation, over pictures of a clean and a dirty wand tip, of why milk left in the steam wand is a hygiene problem, with a transcript available.
• Why everyone gets it: Narration can help when it explains a relevant picture, subject to the content, pace and access needs, as Mayer's multimedia principles describe, without sorting people into "auditory learners".

Read/write: steps and a checklist

VARK's Read/write mode is information displayed as words. It suits rules, steps and anything learners need to check against.

• What VARK counts here: Numbered steps, lists, handouts and checklists.
• In the closing-clean course: The backflush steps written as a numbered list beside the demonstration, and a one-page checklist the barista takes to the machine.
• Why everyone gets it: A checklist is used at the machine, during the task. Written steps also let anyone reading in a second language go at their own pace.

Kinesthetic: a demonstration and real practice

VARK's Kinesthetic mode is a connection to real experience: examples, demonstrations, simulations and practice. It suits procedures and decisions.

• What VARK counts here: A demonstration, a scenario, a simulation or the task itself.
• In the closing-clean course: A short demonstration of a backflush, an activity to put the closing steps in order, a scenario where the cleaning powder container is empty at close, and the supervised clean on shift.
• Why everyone gets it: The close is a physical procedure. Nobody learns to backflush a group head from a diagram alone, whatever their preference.

How to answer a request for VARK-based training

When this request comes up, it usually comes from someone who cares about their people. The aim is to keep their goal and change the method, without making them feel wrong for asking. Here is a short script for the area manager:

1. Start with their goal. "You want every new barista to be able to do the close, whatever way they take things in. That's what I want the course to do too."
2. Share what VARK's own publisher says. "Fleming built VARK as a reflection tool, and VARK's publisher says on its site that it does not advocate matching teaching to learning styles. Studies that give people matched and mismatched versions have found little or no benefit from the match."
3. Offer the design. "So instead of three versions, I'll build one course where each step uses the format it needs: a labelled diagram of the machine, a demonstration with the steps written beside it, a short explanation of why the steam wand must be clean, a checklist for the machine, and practice. Captions and written steps will be there for everyone."
4. Name the measure. "We'll know it worked when new baristas pass the supervised close: right order, no milk in the steam wand, checklist filled in."
5. Keep a place for VARK if they want one. "Anyone curious can take the free official questionnaire for their own study, as a reflection exercise. If we wanted to use it with all staff, we'd need a licence."

The measure in step 4 is a behaviour check on the job, which is level 3 of Kirkpatrick's model. A satisfaction survey that asks whether the course "suited your learning style" would measure the feeling, not the close.

Sometimes the manager will have heard of a program where "VARK worked". Before acting on a claim like that, ask three questions. Did anyone get a mismatched version? Was there a comparison group? And what was measured on the job?

Case study: New laboratory chemists in onboarding training in Egypt

A group of diagnostic laboratories in Egypt trained recently graduated chemists in laboratory parasitology before hiring them, between October 2018 and December 2021. The 100 trainees, aged 22 to 25, needed lab skills before they started work, and the published study credits VARK for the results.

• What they did: Trainees took three online courses of about seven weeks each, delivered through WhatsApp messages and a Facebook page. Everyone got the same content, built to suit two or more VARK modes: labelled images and short videos, narrated commentary, text reviews, and trainees sharing images of their own work. The printed 16-item VARK questionnaire was given at the end of training.
• Evidence and result: In the 2024 BMC Medical Education paper, trainees were grouped afterwards by how many modes they preferred (unimodal 10, bimodal 20, trimodal 40, multimodal 30) and compared on a 13-mark test of case studies, multiple-choice questions and microscope identifications. In the paper's table, mean scores for the trimodal group rose from 2.65 to 12.33 and for the multimodal group from 2.80 to 12.37, while the unimodal and bimodal groups moved from 1.30 to 1.40.
• Limitation: Nobody got a matched or mismatched version, so the study cannot test matching, and there was no comparison group, so the gains cannot be credited to the multimodal design either. VARK was measured after the outcome. The results text and the table report slightly different post-test means. The test was taken at the end of training, not at the bench after hiring.
• Lesson: A real onboarding program that gave everyone the same multimodal content was reported as evidence for VARK. The three questions show that a design like this can describe one program's results, not show that VARK made the difference.

How to design for the content instead of the style

This is the workflow I would use for the café's course, and for any request that starts with learning styles:

1. List the parts of the task. Use what the supervisor checks at sign-off, not a list of topics.
2. Name what each part is. A layout, a procedure, a rule with a reason, a decision or a check.
3. Choose the format the content needs. A diagram for a layout, a demonstration plus written steps for a procedure, a short explanation for a reason, a scenario for a decision and a checklist for a check.
4. Add access options for everyone. Captions and a transcript for any video or narration, written steps beside every demonstration, plain English and alt text on images. These are for access, such as hearing, language or a noisy shop floor, not for a style track.
5. Plan the practice and the check. Each part should be practised in a way close to the real task and checked where the work happens.
6. Measure on the job. Decide before launch what you will observe, such as the supervised close.

Keeping it short suits the setting: a phone-based microlearning course before the first close, then the real practice on shift.

The café's format-by-task plan

Is building separate versions worth the effort? A hospital in Brazil compared advice tailored to each patient's VARK mode with standard advice.

Case study: Tailored dialysis nutrition advice in Brazil

Patients on dialysis often gain too much fluid between sessions. Dietitians at the Santa Casa de Misericórdia de Curitiba university hospital wanted to know whether advice built for each patient's VARK preference would help. These were patients, not staff.

• What they did: In the tailored phase, dietitians built materials for each patient's dominant VARK mode: infographics and flash cards, an oral explanation with an audio message, written guidance with write-in questions, or hands-on food containers and labels, across two sessions of 20 to 30 minutes. In the standard phase, patients got brief spoken guidance without materials. Each phase lasted 30 days, with a break between.
• Evidence and result: In the 2025 randomised crossover trial, 21 patients were analysed. Average weight gain between sessions was 2.53 kg with tailored advice and 2.48 kg with standard advice, a difference that was not statistically significant (p = 0.55). Ten patients were classed as Aural, 3 Read/write, 5 Kinesthetic, 3 bimodal and none Visual.
• Limitation: The sample was small, the phases were short, and fluid gain has many causes. The trial compared tailored with standard advice, not matched with mismatched, and the tailored phase also had far richer materials.
• Lesson: The tailored approach required additional materials and longer counselling, but this small trial did not detect a difference in fluid gain. That supports caution about investing in tailoring; it does not establish equal effectiveness or measure the cost.

How often course briefs mention learning styles

To see whether VARK requests show up in written briefs, I searched 90 days of prompts typed into Coursebox's course builder, in any language, for learning-style phrases in English and six other languages, then read every match by hand to judge whether it described learners by style. We also hand-classified a sample of 104 English-language course requests from the same period, with two readers.

In the closer read of 104 course requests, none asked to let learners choose between formats and none asked for captions or transcripts. Three requests tied a format to the content, such as pictures of correct and incorrect technique or a diagram of a funnel. These counts cover what people wrote in a brief, not what they believe or what their finished courses contained.

The practical reading: VARK requests rarely make it into a written brief, but these prompt data cannot tell us how often they arise in conversations. Access options were absent from the 104 sampled briefs. When a stakeholder asks for VARK versions, the useful thing to write into the brief is the task, the format each part needs and the access options.

VARK compared with VAK, Kolb and other style models

VARK sits in a family of learner-style models, and each one sorts something different. Here is how they compare on what they sort, what they are meant for and what the matching evidence shows. It is my summary, not a ranking.

VAK came first. Barbe and Swassing, with Milone, set out visual, auditory and kinesthetic modality strengths in 1979, and Fleming wrote in 2006 that "VAK inventories had been around for years". His addition was the separate Read/write mode. Kavale and Forness found an effect size of 0.14 across 39 studies that matched reading instruction to modality, as reported in Coffield and colleagues' 2004 review of learning styles for the UK's Learning and Skills Research Centre. That review did not evaluate VARK itself.

Kolb's experiential learning theory is the most useful neighbour for a designer, because the cycle behind his styles is a way to plan practice and reflection, whatever the evidence on the styles. If your aim is to get learners reflecting on how they work, a structured reflection such as Gibbs' reflective cycle gives them more to work with than a style label.

Limitations of the VARK model

VARK is useful as a conversation starter, but these limits shape how far to rely on it:

• It was built as a prompt, not a measure. The 1992 authors chose not to test validity. A 2010 factor-analytic study by Leite, Svinicki and Shi later found "preliminary support for the validity of the VARK scores" and adequate reliability, but flagged problems with item wording and the scoring method, and raised cautions about using VARK in research. Reliable scores on a preference still say nothing about whether matching helps.
• The published statistics come from volunteers. The 1,048,292 respondents chose to take the questionnaire online and skew young, and VARK's FAQ says its database is not necessarily representative of the general population.
• Matching has not been shown to improve learning. Fair tests found no match effect, the VARK-specific study found no link between VARK scores and grades, and the small average effect in the 2024 meta-analysis rests mostly on studies that do not show the crossover.
• Labels can limit people. The Education Endowment Foundation warns that labelling learners can undermine effort, and VARK's own pages say results are not a fixed label and should not be used to assign roles.
• Organisational use needs a licence. Personal use is free. Using VARK with staff, in a paid course or inside your own systems needs permission and usually a fee.
• Its roots are outside experimental psychology. Fleming and Mills drew on neuro-linguistic programming and "left brain, right brain" literature for the starting categories.

Using the VARK evidence with an AI course builder

On 8 October 2026 I briefed Coursebox to build the café's closing-clean course without any learning-style sorting. Coursebox is our product, so weigh my account with that in mind. The brief described the baristas, the five parts of the clean and the supervisor's sign-off, asked for one mobile course of about 25 minutes with each format chosen by what that part of the task needs, and said not to label anything for visual, auditory or kinaesthetic learners or to add a learning-styles quiz.

Most of what I needed came back. There was no learning-style language anywhere in the outline or the course description. The outline had a "Before You Start" section, a "Know the Cleaning Parts" page on the machine parts, a "Backflush Each Group Head" page, a "Put Closing Steps in Order" activity that included the empty cleaning powder scenario, and a "Ready for Sign-Off" section.

Three things did not match the brief. The steam-wand page and the checklist page came back as short quizzes, when I had asked for an explanation with a clean and dirty comparison and a checklist, and there was no wrap-up section.

One follow-up message, asking for those two pages as the brief described them and for a wrap-up, fixed it. The new outline ran to about 36 minutes across 5 sections and 8 pages, longer than the 25 minutes I asked for, and it showed "Keep the Steam Wand Safe" as a clean versus dirty wand page, "Take the Closing Checklist" as a content page, no quizzes, and a final Wrap-up, "What happens at the supervised clean". I checked the outline only, so whether the checklist prints, whether the written steps sit beside the demonstration and whether captions and transcripts exist still need checking in the generated course.

What to do with VARK in your next course

The VARK model describes how people like to take in information, and Fleming built it to help learners reflect on how they study. Use it for that, if at all. Do not use it to build separate versions of a course or to decide who gets which format, because the studies that test matching fairly have found little or no benefit, and neither Fleming nor VARK's publisher proposed sorting learners.

The next time someone asks for training that suits visual, auditory and kinaesthetic learners, sit down with them and fill in a format-by-task plan for that course before anything is built, starting from the goal you share.