AbilityBenchPreference and evidence
Learning style test that then checks its own answer
Sixteen situations produce a four-way preference profile — visual, aural, read/write, kinesthetic — and then a 20-object memory test asks whether your recall actually follows it. Free, no account, no result behind an email. The profile is a genuine measurement of what you prefer; the belief that teaching matched to it improves learning is a separate claim, and it is the one that keeps failing when it is tested properly.
- 100% free
- No signup
- 16 situations
- 20-object recall test
- Profile across 4 modes
Two halves, about eight minutes. First 16 situations with four ways through each, which produces your preference profile — pick as many options per situation as genuinely apply, or none. Then a memory test: 20 everyday objects, 3 seconds each, half of them drawn and half of them printed as the word, followed by writing down everything you remember.
Why the second half exists at all
A preference is a real thing to measure and this page reports yours without hedging. What it will not do is hand you a profile and let you assume the next step, because the next step — that being taught in your preferred mode makes you learn better — is the part that has failed every properly designed test of it. The recall task lets you watch a format effect that does exist, on your own answers, and see whether it lines up with what the questionnaire said about you.
The drawings are line art on a dark panel and each one is announced to a screen reader by name, so a visitor using one is reading the word in both cells and the format contrast does not apply to them — the profile still does. Nothing on this page needs a microphone, a camera or an account.
How to take the learning style test
A questionnaire, a memory list, and the comparison between them.
Answer 16 situations, picking as many options as fit
Each situation offers four ways through it — a diagram, a conversation, a piece of text, a go at the thing itself. Tick every one that is genuinely true of you, which is often two and sometimes none. Multiple selections are how this format is meant to work: a person who would happily read the manual and then take the machine apart is not being indecisive, and forcing a single answer per question invents a preference that was not there.
Watch 20 objects for 3 seconds each, without writing anything
Ten arrive as a simple line drawing and ten as the printed word, assigned fresh each run so the objects are not tied to a format. The first two and last two do not count toward anything — items at the ends of a list are recalled from a different advantage than items in the middle, so they are there as padding. The clock stops if you leave the tab, rather than letting an item sit unwatched and count as study time.
Write down everything you remember, then read the two numbers together
One name per line, in any order, with capitals and plurals ignored and a few obvious alternatives accepted. The result splits your recall into the drawings and the words, sets that difference next to the visual-minus-read gap from your profile, and says plainly what a single 16-item run can and cannot support. The full study list is shown afterwards, so you can check every item the matcher scored.
Technical specifications
| Inventory | 16 situations, 4 options each, multiple selections permitted and blanks counted as blanks rather than as zeros. The items are original to this page: the published questionnaire's wording belongs to its owner and none of it is reproduced, which is also why a profile here is not a score from that instrument |
|---|---|
| Scoring the profile | One point per selected option, reported as four counts out of 16 with no conversion into a type, a band or a percentage of a population. A tie at the top is reported as a tie, and a lead of one selection is flagged as inside the noise of a questionnaire this short |
| Recall list | 20 everyday objects, 3 seconds each, 60 seconds of study in total. 16 are scored — 8 drawings and 8 printed words — and 4 sit in the first two and last two positions as buffers against primacy and recency |
| Which two modes are tested | Visual against read/write, and only those two. Speech would have to come from whatever voice the machine carries, so an aural cell would be measuring the operating system; a kinesthetic cell means doing the thing, which a page cannot arrange. Two of four is a limit stated rather than faked |
| Recall matching | A typed name counts when it matches the object or a listed alternative after case and a trailing s are removed — cup takes mug, boat takes ship, envelope takes letter. Names matching nothing are counted separately as intrusions rather than ignored |
| Drawings | Inline SVG line art, no image files and nothing downloaded. Each drawing carries the object's name for a screen reader, which means a visitor using one reads the word in both cells — the profile still applies to them, the format contrast does not, and the page says so up front |
| Reference position | No norm and no percentile: there is no population distribution for this list of objects, and no defensible reference for a preference count. The relevant literature is a finding rather than a figure — Massa & Mayer (2006), Testing the ATI hypothesis: should multimedia instruction accommodate verbalizer-visualizer cognitive style?, Learning and Individual Differences |
| What leaves the page | Nothing. The recall matching runs against a list this tab generated a minute earlier, your typed text never leaves the tab, and the copy button writes plain text to your own clipboard |
Frequently asked questions
So is my learning style real or not?
Your preference is real; the instruction that is supposed to follow from it is not. Two different claims travel under one name, and separating them is most of what this page is for. People do reliably report preferring diagrams over text, or discussion over reading, and those reports are reasonably stable over time — that is a measurable thing about a person. The second claim is that matching how you are taught to that preference improves what you learn, and that one needs a specific result: learners of style A must do better under method A and worse under method B, with the reverse for learners of style B. Reviews that went looking for that crossing pattern in the literature found almost no studies designed to detect it, and the well-designed ones that exist did not find it.
Why does the memory test only use pictures and words?
Because those are the two conditions a browser can present without measuring the machine instead of you. An aural condition means synthetic speech, and the voice, the rate and the clarity vary so much between operating systems that half the visitors would be studying under a different stimulus from the other half — the aural cell would be a measurement of somebody's voice pack. A kinesthetic condition means physically doing the thing, and there is nothing here to do. Rather than invent two cells and quietly weaken all four comparisons, the page runs the two it controls exactly and says which two are missing.
The drawings beat the words for me. Doesn't that make me a visual learner?
It makes you a person, because that is what nearly everybody does on this list. Pictures are recalled better than words in free recall as a matter of routine, and the effect turns up in people whose questionnaire says read/write just as readily as in people whose questionnaire says visual. That is precisely the trap the page is built around: a real, large, reproducible format effect exists, it feels like confirmation, and it is not evidence of matching because it happens to everyone in the same direction. The number worth watching is not which cell won but whether the size of your advantage tracks the size of your preference — and with eight items a cell, on one person, it cannot.
Why are the first two and last two items thrown away?
Because where an item sits in a list changes how likely you are to recall it, independently of how it was shown. The opening items get rehearsed against an empty list and end up better remembered; the closing items are still sitting in immediate memory when the recall starts. If those four positions were scored, and the format assignment happened to put three drawings among them, part of the reported difference would be list position wearing a format label. Padding the ends is the cheap standard fix, and it costs four items out of twenty.
I typed something that was not on the list. What happened to it?
It was counted as an intrusion and shown to you as one. False recall is a real phenomenon rather than carelessness — a study list of everyday objects reliably produces confident memories of related things that were never presented — so the number is reported instead of silently discarded. The other possibility is more mundane: the matcher accepts a fixed set of alternatives per object, and a reasonable synonym nobody anticipated will be counted as an intrusion too. The full list is printed with the result so you can see which of the two happened.
Can I use this result to ask for different teaching?
You can use it to say what you prefer, which is worth saying, and not to claim it will make you learn more. The distinction matters in practice because instruction designed around a supposed style has a cost: time spent producing four versions of the same material is time not spent on the things that do move learning, like spacing the practice out, testing yourself instead of rereading, and getting worked examples early. The material's own best form is usually the one to argue for. A map is better as a map for everyone in the room.
Is this the official VARK questionnaire?
No, and none of its items are reproduced here. The situations on this page were written for it, the option wording is ours, and the scoring is a plain count with no bands, so a result here is not comparable to a result from the published instrument and should not be reported as one. The mark is used only to name the four-way model this page follows, and the non-affiliation line below says so explicitly.
Where the four letters came from, and where the evidence stops
The model is younger and narrower than its reputation. Four modes — visual, aural, read/write, kinesthetic — were proposed in the early nineties as a way of getting university teachers to think about how they were presenting material, with the questionnaire framed as a conversation starter rather than a measuring instrument. The split of visual from read/write is the model's one genuinely useful move, because most earlier schemes had folded reading in with seeing and thereby made the commonest preference invisible. What happened afterward is the familiar arc: a reflective prompt for teachers became a label attached to learners, and the label acquired a promise nobody had tested.
That promise has a technical name — the meshing hypothesis — and a specific shape, which is why it is testable and why its failure is not a matter of opinion. To show that matching works you need an interaction, not an effect: it is not enough that everybody learns well from pictures, or that visual learners score highly. You need the visual group to do better under the visual method and worse under the verbal one, while the verbal group does the reverse. When that criterion was applied to the published literature, the striking finding was how few studies had even been designed to detect it, and how consistently the ones that were came back empty. The pattern has held in the direct tests since, including work that classified people by preference and then taught all of them both ways.
Which leaves the more interesting question of what a preference is a measurement of. It is not nothing: something stable is being reported, it just is not a channel your brain learns through. Compare it with the differences on this site that do behave like traits — the imagery a person actually has varies from photographic to entirely absent, and unlike a style preference it predicts performance on tasks that require imagery. Or the sensory profile, where a difference in how much input somebody wants is measured per modality because the modalities genuinely do not move together. The aural column is the one where that distinction bites hardest: liking a lecture is a taste, while a strong physical reaction to particular sounds is not one, and the questionnaire above cannot tell them apart. A preference for diagrams sits closer to a taste than to any of those, and taste is worth knowing without being worth prescribing from. The same substitution of a label for a measurement is what hemisphere dominance does to lateralization, and what time blindness does to interval timing. If it is speed of processing you are actually curious about, that has its own measurement on the processing speed test.
Why no benefit figure is printed anywhere on this page
The claim that teaching matched to a learner's preferred style improves outcomes is the thing the inventory is used for, and it is unsupported: the reviews that looked for the required interaction effect did not find it. There is therefore no norm because there is no effect to norm.
Pashler, McDaniel, Rohrer & Bjork (2008), Learning styles: concepts and evidence, Psychological Science in the Public Interest
The two direct tests this page leans on: Massa & Mayer (2006), Testing the ATI hypothesis: should multimedia instruction accommodate verbalizer-visualizer cognitive style?, Learning and Individual Differences. The model itself: Fleming & Mills (1992), Not another inventory, rather a catalyst for reflection, To Improve the Academy.
This is a measurement exercise, not a clinical assessment. It reports what you did on this page against a stated reference and nothing more — it cannot establish dyslexia, a language difficulty, or anything at all about how you ought to be taught. Only a qualified professional, working with more than a browser, can make that judgment.
Not affiliated with, endorsed by, or connected to the owner of VARK. VARK is a trademark of its owner and is used here only to name the assessment this page prepares for. The questions on this page are our own: no part of the published test is reproduced, and a score here is not comparable to a score from the real instrument.
Where your answers and your recall list are handled
Every number on this page is worked out by JavaScript running in the tab you are reading it in. Your answers, your reaction times and your score are never uploaded, logged or kept — which is also why the test carries on working after you disconnect from the network, and why nothing here can be held back behind an email address.
The study list is generated in this tab from a seed drawn when you start, and the text you type into the recall box is matched against that list here, in the page. Neither the questionnaire answers nor the recalled names are written to storage or sent anywhere, so reloading loses the run — which is why the seed is included in the copyable summary.