HSW-0275 · How Studying Works
“I’m a visual learner.”
“I only remember things if I hear them.”
“I learn best by doing.”
These statements may describe real preferences. A learner may genuinely enjoy diagrams, dislike long lectures, prefer to talk through a problem or feel more engaged when physically manipulating something.
The error begins one step later: assuming that a preferred format must therefore produce the best learning.
Preference is evidence about experience. Performance is evidence about learning. They can agree, but they are not the same variable.
This article owns that gap. It does not argue against personalisation; use How Personalised Learning Works for the broader question of adapting routes while preserving the learning destination. It does not own strategy selection generally; use How Learning Strategies Work. Here the job is narrower: how should a student use preference without mistaking it for proof of effectiveness?
Quick Answer
The classic learning-styles claim is a matching hypothesis: identify a learner’s preferred modality or style, then teach that learner in the matched mode because the match should improve achievement. Evidence for that strong claim remains weak.
A 2024 meta-analysis found an overall small positive effect for matched instruction across 21 eligible studies, 101 effect sizes and 1,712 participants (g = 0.31). But the pattern required to support the matching hypothesis—a crossover where different learner groups benefit from different matched methods—appeared in only about 26% of the analysed learning outcomes. The authors judged the benefits too small, infrequent and supported by studies of too-low quality to warrant widespread adoption. See Clinton-Lisell and Litzinger, 2024.
A 2025 Educational Psychology Review analysis reached a related conclusion from a different angle. Across meta-analyses, studies that directly tested matching produced effects near zero, while correlational studies often mixed together learning styles, preferences and learning strategies. The authors argue for shifting attention from fixed style labels toward adaptable strategies selected for the learning goal and task. See Hattie and O’Leary, 2025.
The practical conclusion is not “preferences are fake.” It is: preferences should inform access, comfort, motivation and choice; delayed performance should decide whether the study method works.
Like is not learn. Dislike is not fail. Test the capability after the support, format and feeling have changed.
1. Three Different Questions Are Often Collapsed Into One
- Preference: How do I like information to be presented or practised?
- Strategy: What cognitive operation am I using—retrieval, explanation, comparison, worked examples, practice problems?
- Performance: What can I understand, remember, select or do later?
A learner can prefer a weak strategy. A learner can dislike an effective strategy because it feels difficult. A learner can also prefer a format that genuinely improves access. The three questions need to stay separate long enough to discover which is happening.
2. Preferences Are Psychologically Real
People have likes and dislikes. Those preferences can affect willingness to start, persistence, emotional comfort and whether a learner uses a resource voluntarily.
There is no need to deny that. The evidence problem concerns the next claim: whether matching instruction to a self-reported style reliably produces better learning than a non-matched method.
Keeping this distinction prevents an unhelpful argument in which rejecting the matching hypothesis is misheard as rejecting learner differences altogether.
3. What a Real Matching Test Requires
Suppose students are classified as visual or auditory. A strong test of the meshing hypothesis cannot simply show that visual learners scored well after a visual lesson.
It needs an interaction: visual-preferring learners should learn better from the visual method than the auditory method, while auditory-preferring learners should show the opposite pattern. Learners need comparable assessment conditions, and assignment to learning methods should be controlled well enough to support the inference.
This crossover requirement is why many ordinary correlations do not answer the matching question.
4. Correlation Is Not Matching
If students who prefer analytical study tend to earn higher marks, several explanations are possible.
- The preference may cause better learning.
- More capable students may have learned to prefer effective strategies.
- The preference measure may partly measure strategy use rather than a stable style.
- A third variable—prior knowledge, motivation, course design—may influence both.
That is why Hattie and O’Leary’s 2025 analysis separates direct matching studies from correlational studies. The two answer different questions and should not be pooled into one story.
5. The 2024 Meta-Analysis Is More Nuanced Than “Learning Styles Debunked”
Clinton-Lisell and Litzinger’s 2024 meta-analysis is useful precisely because it does not report a simple zero. The average matched-versus-unmatched effect was positive and statistically significant.
But a matching theory requires more than an average benefit. If one method tends to be generally better for most students, matching a subgroup to it does not prove that each subgroup needs a different method. The authors therefore examined crossover interactions. Only about a quarter of learning-outcome measures showed the pattern required for at least two styles.
They also reported substantial heterogeneity and low study quality. Their conclusion was not that no study has ever found a matching benefit. It was that the evidence is too inconsistent and costly to justify widespread style-matching practice.
6. The 2025 Reanalysis Explains Why the Literature Looks Confusing
Hattie and O’Leary reviewed 17 meta-analyses and argued that two different literatures had been blurred together.
Meta-analyses of direct matching studies produced an overall effect near zero. Correlational meta-analyses showed modest relationships between style-like measures and achievement, but these studies did not test whether matching instruction to preference caused better outcomes.
The broader lesson for students is important: a label may be describing a strategy tendency, a preference, a habit, a self-concept or a real capability difference. Those are not interchangeable.
7. Content Has a Best Representation Sometimes
Rejecting learning-style matching does not mean every representation is equally good for every idea.
A map is often better than a spoken paragraph for showing spatial relationships. A graph can reveal a trend that a list of numbers hides. Pronunciation requires auditory information. Dance technique needs movement. Algebraic structure benefits from symbolic representation.
The important reason is the structure of the content and the target capability, not the claim that one learner is permanently “visual” while another is permanently “auditory”.
8. Accessibility Is Not a Learning Style
A learner with low vision may need audio or enlarged text. A student with a hearing impairment may need captions or sign language. A learner with dyslexia may use text-to-speech or adjusted typography.
These are access needs and accommodations, not weak evidence for the learning-style matching hypothesis. Preserve legitimate supports. Test learning within the learner’s appropriate access conditions.
9. Preference Can Improve Adherence Without Improving Learning Per Minute
A student may prefer video and therefore spend thirty minutes willingly watching an explanation that they would abandon after five minutes of reading.
That preference can matter because no method works if the learner never uses it. But the correct outcome question remains: after the video, what can the learner explain or do without the video?
Preference may improve entry and persistence while a different operation—retrieval, problem solving, explanation—produces the durable learning.
10. Fluency Can Masquerade as Fit
Preferred formats often feel fluent. A familiar narrator, clean diagram or favourite app reduces friction.
Fluency can be helpful when it removes irrelevant difficulty. It can also produce overconfidence when the learner interprets easy processing as strong memory.
A study method should therefore be judged after support is removed and time has passed, not only during the pleasant experience of using it.
11. Desirable Difficulty Can Make Good Methods Feel Worse
Retrieval practice can feel harder than rereading. Interleaving can feel less orderly than blocked practice. Explaining a concept from memory can feel less comfortable than highlighting a polished summary.
If preference is allowed to decide every study move, learners may systematically avoid methods that reveal forgetting and uncertainty.
The solution is not to make studying unpleasant. It is to distinguish useful effort from needless friction, then let delayed performance arbitrate when feeling and learning disagree.
12. Preference Can Be a Starting Constraint, Not the Final Verdict
Suppose a learner strongly prefers spoken explanations.
Use that preference to enter the concept. Then require a changed operation: draw the causal model, answer questions without replaying the audio, explain it in writing, solve a new case.
Now the preference helps access while transfer prevents the learning from becoming trapped in one format.
13. Mathematics: The Task Decides the Representation
A student may say they prefer videos for Mathematics. A worked video can be useful for seeing a method unfold.
But the examination will eventually require the student to recognise the problem structure, choose a method and execute it without the narrator. Therefore, the study sequence should cross formats: watch or inspect a worked example, close it, reconstruct the method, solve a changed problem, then explain why the method applies.
The preferred format starts the route; independent problem solving validates it.
14. English: Listening Preference Does Not Remove the Need to Read
A learner can prefer audiobooks and still need to develop close reading of printed or digital text. Audio may improve access to plot, vocabulary pronunciation and expressive phrasing. It does not automatically train every visual reading demand: punctuation, paragraph structure, evidence location or silent exam reading.
Use audio where it contributes. Then practise the modality required by the target capability.
15. Science: Diagrams Are Powerful Because the Content Is Relational
A learner who likes diagrams may genuinely benefit from them in Science—but not because a questionnaire discovered a visual essence.
Diagrams can externalise spatial and causal relations: blood flow, circuits, forces, particle arrangements. The representation fits the knowledge. After using it, the learner should still explain the mechanism without simply pointing at arrows.
16. “Hands-On” Can Mean Several Different Things
A student may prefer doing rather than listening. That can refer to experimentation, physical manipulation, immediate feedback, autonomy, movement or simply relief from passive instruction.
Before classifying the learner as kinesthetic, ask what property of the activity improved engagement or understanding. The correct repair may be active problem solving rather than movement itself.
17. The Preference Audit
- Name the preferred study format or activity.
- State what you like about it: clarity, pace, control, social contact, sensory form, feedback, low anxiety?
- Identify the cognitive operation inside it.
- Predict what you will remember or do after delay.
- Test the same target with an alternative method.
- Compare delayed performance, not just immediate comfort.
This turns preference into a testable design variable rather than an identity label.
18. Use Common Outcomes When Comparing Methods
If Method A is tested with pictures and Method B with words, the assessment may favour one by resemblance alone.
To compare learning, use a common target where possible: delayed explanation, recall, application or transfer. This is the same logic behind careful tests of learning-style matching.
A preference claim becomes more credible when it predicts performance under a fair, shared outcome measure.
19. The Study–Test Format Relationship Is Different
Practising in a format similar to the final test can sometimes help because cues, procedures or response demands overlap. That is not evidence that the learner has a stable style.
Use Study–Test Format Match for that narrower problem. A learner preparing for an oral examination should practise speaking. A learner preparing for a handwritten paper should practise producing answers under that constraint. The reason is task alignment, not personality classification.
20. Personalisation Should Adapt the Route, Not Shrink the Destination
Good personalisation can vary pacing, examples, scaffolds, order, accessibility, feedback and practice dose.
It should not quietly tell a learner, “You are not a reading person, so you never need to read,” or “You are visual, so you do not need to explain in words.” Those labels can become capability ceilings.
The destination should remain the capability the subject actually requires.
21. Parent and Tutor Guide: Replace Labels With Observations
Instead of “She is a visual learner,” try:
- She begins faster when a diagram shows the whole system.
- She remembers labels but loses causal relationships after a day.
- She avoids closed-book retrieval because it feels uncomfortable.
- She understands spoken explanation but struggles to reconstruct it in writing.
- She performs better when examples are compared side by side.
These statements can be tested and repaired. A global identity label is much harder to use responsibly.
22. A Two-Week Preference–Performance Experiment
Choose two comparable topics.
- Study Topic A using the learner’s preferred method.
- Study Topic B using a well-supported alternative matched to the task.
- Keep time approximately equal.
- Record enjoyment, perceived effort and predicted score immediately.
- Test both after forty-eight hours without study materials.
- Test one transfer item for each topic.
- Reverse the methods with two new topics the following week.
This is not a laboratory experiment. The topics will never be perfectly matched. But repeated within-person comparison can reveal whether preference, prediction and delayed performance are moving together or apart.
23. When Preference Should Carry More Weight
- When several methods produce equivalent delayed performance.
- When the preferred route substantially improves adherence.
- When the choice is about presentation rather than the target operation.
- When accessibility or sensory needs make one route genuinely more usable.
- When motivation would otherwise collapse and the preferred entry preserves engagement without reducing the final capability.
If two routes teach equally well, there is no prize for choosing the one the learner hates.
24. When Preference Should Carry Less Weight
- When delayed tests repeatedly show weaker retention.
- When the preferred method removes the operation the examination requires.
- When it produces familiarity but not independent retrieval.
- When it encourages avoidance of a capability the subject genuinely needs.
- When the learner cannot transfer beyond the preferred format.
25. Evidence Boundary
Learning-styles research contains many definitions, inventories and study designs. The 2024 meta-analysis found a small positive average effect, so it would be inaccurate to say that every matching study is null. The stronger problem is that the specific crossover pattern required by the theory appeared infrequently, study quality was low, and results were highly variable.
The 2025 Hattie–O’Leary paper is a reflection and meta-analytic reinterpretation, not a new randomized classroom trial. Its contribution is conceptual discrimination: matching studies and correlational style studies should not be treated as the same evidence.
None of this implies that all learners are identical. Learners differ in prior knowledge, disability, language, motivation, strategy repertoire, self-regulation and experience. The evidence simply does not justify converting ordinary modality preferences into a universal rule for matching instruction.
26. Return: Let Preference Open the Door; Let Performance Judge the Route
Preferences matter because learners are human.
But a preference is not a certificate of cognitive effectiveness.
Use preferred formats when they improve access, willingness and clarity. Choose representations because they fit the knowledge. Choose strategies because they train the target operation. Then wait, change the cues, remove the support and see what the learner can still do. That is where preference becomes accountable to learning.
Continue through How Learning Strategies Work, How Personalised Learning Works, Study–Test Format Match, the How Studying Works Numbered Series Reading Index and the How X Works Hub.