learning styles: a seductive but utterly useless idea

Fuggetaboutit, they don't exist

‘Learning styles’ is simple in theory, possibly even flattering to the learner, and utterly ridiculous in practice (when you think about it carefully).

(This is a cross-post from my BrainPizza Newsletter (S'ouvre dans une nouvelle fenêtre) - I’m just trying out this platform as an alternative)

The theory: if we somehow match teaching to a learner’s preferred modality (typically labelled ‘visual, auditory, reading/writing, or kinaesthetic’, sometimes shortened to VARK), learning will improve (somehow).

This is a big claim. Imagine being confronted with a task to learn: a written script (verbal learning), or the piano (visuomotor learning), or navigating a complex work situation with competing demands (social learning), or learning a song (auditory); or the new complex ticket machine at your train station (visual, auditory, kinaesthetic); or learning to drive a car, a bike or a computer game (visual, auditory, reading, kinaesthetic).

Transform the task somehow to your preferred learning style, and voilà, you’ll learn more quickly, efficiently, and retain the material for longer, and presumably be able to use it more effectively.

On the contrary, if you ignore your preferred learning style, then you’ll learn more slowly and less effectively.

Seductive idea, isn't it? (First question: how do you know your preferred learning style? Do you really know it? On reflection, do you have a preferred style for learning how to swing a racket, a golf club, or to learn a new complex coding procedure? Course you don’t.)

Learning styles theory is not just a claim about some preference you have (like a preference for salty over sweet foods); it is a serious claim about the mechanisms of learning and memory, as well as a set of predictions about the learning and memory systems of the brains of learners.

using your memory to strengthen your memory: retrieval practice Shane O'Mara (S'ouvre dans une nouvelle fenêtre) Read full story (S'ouvre dans une nouvelle fenêtre)

However, pursuing learning styles as a theory of learning and memory has been one of the deadest of dead ends: a pointless, nonsensical, take on learning and memory, a view of learning discussed solely within a group of aficionados who have been having silly and pointless conversations with themselves for decades, even though the idea has failed every empirical test, and makes no contact with actual studies of learning and memory, at any level, from synapses to brain systems to collective memory.

interleaving: the mix-it-up memory method that works Shane O'Mara (S'ouvre dans une nouvelle fenêtre) Read full story (S'ouvre dans une nouvelle fenêtre)

And it is a daft idea if you step back and consider the anatomy and evolved cognitive functions of the brain in any detail at all:

Where is my memory? Shane O'Mara (S'ouvre dans une nouvelle fenêtre) Read full story (S'ouvre dans une nouvelle fenêtre)

In experimental terms, learning styles proponents assert an aptitude–treatment interaction (ATI): a person × treatment effect in which people assigned to their preferred ‘style’ will learn more from instruction delivered in that style, and will learn less when paired to a mismatched style.

Sins of Memory Shane O'Mara (S'ouvre dans une nouvelle fenêtre) Read full story (S'ouvre dans une nouvelle fenêtre)

That decisive crossover pattern has not materialised - for it does not exist. Instead, we have an intuitively-appealing story floating free of neurocognitive theory, resting on little-to-no or weak measurement, and which resists being killed off when critical tests show it is wrong (because the committed want to believe in it!).

My book ‘Talking Heads (S'ouvre dans une nouvelle fenêtre)’ (US Kindle Ed available here (S'ouvre dans une nouvelle fenêtre)) has an extensive discussion of shared reality, memory, and collective remembering. No learning styles here.

The prevalence of belief in learning styles (S'ouvre dans une nouvelle fenêtre) among educators remains strikingly and stupidly high, possibly explaining the persistence of the learning style neuromyth (S'ouvre dans une nouvelle fenêtre) despite null findings.

The effective alternatives (retrieval practice (S'ouvre dans une nouvelle fenêtre), spacing and interleaving (S'ouvre dans une nouvelle fenêtre))are rooted in well-replicated cognitive mechanisms that generalise across learners.

Finally, contemporary neuroscience shows memory depends on distributed, interacting cortical–temporal–diencephalic circuits (S'ouvre dans une nouvelle fenêtre), and not isolated ‘visual’ or ‘auditory’ modules, undermining the biological plausibility of fixed sensory ‘styles.’

If you enjoyed this post, and want more essays on how neuroscience and psychology can help us build better lives, subscribe below. You’ll get new ideas from BrainPizza (S'ouvre dans une nouvelle fenêtre).

the seduction of personalisation

Learning styles continue as a conjecture about learning because they take some very weak claims, which are then taken to a stronger set of mistaken conclusions.

It is true that disciplines traffic in different representational media, from maps and diagrams in geography, to equations and graphs in physics, to timelines and narratives in history. And multi-modal materials for instruction may help learners.

From these premises, a stronger claim is made: that the durable part of learning improves when instruction is matched to a person’s stable modality preference.

The rhetoric is that of personalisation, but the reality is a label applied to people rather than an analysis of the tasks they face.

A credible theory of learning makes contact with mechanisms and boundary conditions. It specifies what is changing in the learner (encoding, storage, retrieval); it says when a predicted effect should appear and when it should not.

Learning styles accounts offer none of this. ‘Preference’ is treated as a cause; ‘matching’ is treated as a cure, leaving us to imagine a ‘visual learner’ who ought to learn better from pictures and an ‘auditory learner’ learning geometry by talk rather than diagrams (as if they don’t have a visual system!).

This is not a theory; this is ‘not even wrong’.

a test the idea cannot pass

The only fair test are experiments that:

(i) classify learners with a stable, valid measure;

(ii) randomly assign them to matched or mismatched treatments;

and (iii) evaluate outcomes that matter (delayed retention and transfer), while estimating the crossover interaction (S'ouvre dans une nouvelle fenêtre) directly.

(BTL: what modern cognitive science actually says; neuroscience says not to styles; harms and opportunity costs; why the myth persists; from labels to learning architecture; policy and practice: what to stop, what to start; the bottom line; annotated bibliography, with brief notes on methods and relevance)

No learning style effects emerge from such tests.

Where effects appear, they are typically treatment main effects that help everyone: well-designed diagrams improve spatial reasoning; focused listening improves phonological awareness; hands-on practice improves motor skill.

Where style inventories predict anything, it is usually preference or satisfaction, not durable learning.

Measurement compounds the problem with learning styles:

• Learning style inventories show poor test–retest reliability, high inter-correlations among ‘styles,’ (they should be orthogonal!), and close to zero predictive validity once prior knowledge, general ability, and motivation are taken into account.

• The learning style theory’s response is not been cumulative refinement, but taxonomic proliferation (VARK one year, Kolb or Felder–Silverman the next), all accompanied by post-hoc redefinitions of success (engagement, confidence) when learning gains are not there.

• In Lakatos’s terms, this is a degenerating research programme (S'ouvre dans une nouvelle fenêtre): anomalies do not lead to sharper, riskier predictions but to ever more elastic interpretations.

• And yet this nonsense continues to be perpetuated (S'ouvre dans une nouvelle fenêtre) (search ‘learning styles (S'ouvre dans une nouvelle fenêtre)’ for more of this stuff (S'ouvre dans une nouvelle fenêtre)).

This Substack is reader-supported. To receive new posts and support my work, consider becoming a free or paid subscriber.

what modern cognitive science actually says

Although learning styles theory hypes a non-existent phenomenon, there is an alternative - the modern science of learning.

Cognitive psychology has mapped the constraints and levers that matter: limited attention and working memory, the benefits of well-aligned words and images (dual coding) when they explain rather than decorate, the power of retrieval practice for strengthening memory traces, and the reliable gains from spacing and interleaving that improve discrimination and retention.

These are general mechanisms that apply to everyone, modulated by task demands and prior knowledge, not by a person’s declared sensory preference.

Crucially, when modality matters, it matters because of content, not person. Some material is intrinsically spatial or diagrammatic; some is temporal or phonological; some is symbolic and best learned through worked examples and practice problems. The pertinent question is, “Which representation reveals the structure of this idea?”, not, “Which box did the learner tick on a self-report inventory?”

neuroscience says no to styles

The neuroscience of memory likewise points away from modality labels and toward the architecture of representation and retrieval. Human memory is supported by an interdependent, tripartite circuit linking cortical temporal regions (e.g., perirhinal cortex and temporal neocortex) that compute item-level representations and support familiarity; the hippocampal formation that binds items to space and time into relational episodes via pattern separation and completion; and diencephalic anterior thalamic nuclei, with mammillary bodies and retrosplenial connections, coordinating hippocampal–cortical communication and temporal–context signals. Lesions anywhere along this circuit produce amnesia-like deficits that cut across sensory modality (learning styles do not predict this).

What distinguishes successful from unsuccessful learning is not whether information arrived via eye or ear, but whether the system was afforded the representations and retrieval conditions it needs: clear relational structure, spaced practice, and opportunities to reinstate context.

From this vantage, matching instruction to a modality preference is orthogonal to the underlying machinery. Instruction helps when it aligns form to function (when diagrams present spatial relations, or when examples make abstract schemas concrete, or when practice retrieves and recombines knowledge under varying cues).

harms and opportunity costs

It’s tempting to treat learning styles as harmless: if it motivates teachers to vary materials and students to engage, where is the damage?

Here’s the problem:

First, once students are labeled (‘I am an auditory learner’), the label can harden into identity and avoidance. A learner who most needs to become competent with graphs or equations now has a rationale for steering away from them.

Second, time and money are finite. Every hour allocated to administering inventories, re-labeling materials, and chasing matches is an hour not spent on practices with demonstrable effects:

• building retrieval into courses;

• spacing reviews over weeks;

• interleaving problem types;

• aligning worked examples to likely misconceptions;

• and giving informative, timely feedback.

These work, btw - there’s a vast body of data showing this.

why the myth persists

The persistence of learning styles has less to do with evidence and more with psychology and markets. Broad, flattering categories feel personal (the Barnum effect). ‘Personalisation’ carries a strange political and commercial appeal: parents like it; publishers can package it. And good practices (ulti-modal materials; optionality in how students demonstrate understanding) are routinely, and wrongly, attributed to style matching. In short, success is misdiagnosed and credit misassigned.

from labels to learning architecture

There is a better way to personalise learning: adapt not to a declared modality, but to what actually matters:- readiness, prior knowledge, misconceptions, and rates of progress - the most productive personalisation is the unglamorous work of aligning tasks to goals and supports to needs.

That means teaching the disciplinary literacies a field demands (reading graphs, writing explanations, manipulating symbols, constructing diagrams), choosing representations revealing a concept’s deep structure, and designing practices that forces retrieval under varied cues, conditions, and delays.

policy and practice: what to stop, what to start

Institutions should stop administering style inventories, and cease promising staff or parents matching will raise attainment.

Instead: invest in staff development on the cognitive principles with the strongest support:

• how to build retrieval into weekly rhythms;

• how to space and interleave effectively;

• how to align words and images to explain rather than decorate;

• how to diagnose and address common misconceptions;

• how to calibrate students’ metacognition so they can plan, monitor, and evaluate their learning with greater accuracy.

Evaluate success not by immediate liking or perceived ease, but by delayed retention and transfer to novel problems: outcomes signalling genuine learning.

the bottom line

Learning styles takes a small insight (presentation of to be learned matters, a tiny bit, for reasons unrelated to learning) and attaches it to an aptitude–treatment interaction the evidence does not support. Moreover, the neuroscience of memory offers no mechanism by which a stable sensory preference governs durable learning across tasks; and the programme has failed to generate sharper predictions in the face of null results.

Learning styles: fuggetaboutit, they don't exist.

annotated bibliography (with brief notes on methods and relevance)

Pashler, H., McDaniel, M., Rohrer, D., & Bjork, R. (2008). Learning Styles: Concepts and Evidence. Psychological Science in the Public Interest, 9(3), 105–119. (S'ouvre dans une nouvelle fenêtre)A definitive methodological review: they specify the experimental design required to validate ‘matching’ (an aptitude–treatment interaction with a disordinal crossover) and show nearly all extant studies either don’t test this criterion or fail it. Conclusion: no credible evidence that tailoring by putative “style” improves learning.

Coffield, F., Moseley, D., Hall, E., & Ecclestone, K. (2004). Learning styles and pedagogy in post-16 learning: A systematic and critical review. Learning and Skills Research Centre (LSRC). (S'ouvre dans une nouvelle fenêtre)Finds weak reliability/validity, conceptual incoherence, and warns against policy/practice built on such measures.

Rogowsky, B. A., Calhoun, B. M., & Tallal, P. (2015). Matching learning style to instructional method: Effects on comprehension. Journal of Educational Psychology, 107(1), 64–78. (S'ouvre dans une nouvelle fenêtre) A pre-registered, PSPI-compliant test of the matching claim: participants classified as ‘auditory’ or ‘visual,’ randomly assigned to audiobook vs. e-text, then tested immediately and after delay. No aptitude–treatment interaction and no benefit from matching preference to modality.

Rogowsky, B. A., Calhoun, B. M., & Tallal, P. (2020). Providing instruction based on students’ learning-style preferences does not improve learning. Frontiers in Psychology, 11, 164. (S'ouvre dans une nouvelle fenêtre)An expanded replication/extension consolidating the 2015 null results and situating them within the broader literature; again, no instructional advantage from aligning to self-reported style.

Husmann, P. R., & O’Loughlin, V. D. (2019). Another nail in the coffin for learning styles? Anatomical Sciences Education, 12(1), 6–19. (S'ouvre dans une nouvelle fenêtre)In a large anatomy cohort (n≈426), students’ VARK categories did not relate to chosen study strategies or grades; some techniques helped all students, but gains were unrelated to any learning style ‘match.’ Reinforces that preferences ≠ optimized learning pathways.

Willingham, D. T., Hughes, E. M., & Dobolyi, D. G. (2015). The scientific status of learning-styles theories. Teaching of Psychology, 42(3), 266–271. (S'ouvre dans une nouvelle fenêtre)A concise review clarifying why learning-styles claims lack predictive power and why educators should instead focus on task demands, prior knowledge, and evidence-based strategies.

Newton, P. M. (2015). The learning styles myth is thriving in higher education. Frontiers in Psychology, 6, 1908. (S'ouvre dans une nouvelle fenêtre)Text-mines education databases and shows that the majority of papers referencing ‘learning styles’ endorse or assume their utility, despite the lack of evidence, illustrating how the myth sustains itself.

Newton, P. M., & Salvi, A. (2020). How common is belief in the learning-styles neuromyth and does it matter? A pragmatic systematic review. Frontiers in Education, 5, 602451. (S'ouvre dans une nouvelle fenêtre) Systematic review of educator belief: across 37 samples in 18 countries (N≈15,405), about 89% endorse matching—belief far outpaces evidence; brief corrective interventions can reduce belief substantially.

Aggleton, J. P., & O’Mara, S. M. (2022). The anterior thalamic nuclei: core components of a tripartite episodic memory system. Nature Reviews Neuroscience, 23, 505–516. (S'ouvre dans une nouvelle fenêtre)Proposes a tripartite episodic memory system—temporal (hippocampal) and medial diencephalic (anterior thalamic) streams that converge on shared cortical areas—integrating lesion, tract-tracing, and physiology. Provides the neuroscientific scaffold that learning relies on distributed circuits rather than single-modality ‘styles.’