Spatial Learning: What VR Really Does for Students

Overview

Almost every argument for virtual reality in the classroom eventually cites the same statistic: we remember 10% of what we read and 90% of what we do. It appears in conference decks, product brochures and education journalism, usually credited to Edgar Dale's Cone of Experience.

Dale never wrote those numbers. He never attached percentages to his cone at all. The figures were invented, and they have been traced to an employee of Mobil Oil who published them in a 1967 trade magazine article — from which they were copied, uncredited, into decades of training literature.

This matters because the real evidence for VR learning is genuinely strong, and it does not need a fabricated statistic propping it up. Peer-reviewed meta-analyses covering hundreds of studies and tens of thousands of participants find substantial measured benefits.

They also find something the 90% claim obscures entirely: VR works dramatically better for some subjects than others, and the size of the advantage shrinks considerably once you account for how well the underlying studies were designed.

Where the 10% and 90% Numbers Came From

The full version of the myth is a tidy ladder: 10% of what we read, 20% of what we hear, 30% of what we see, 50% of what we see and do, 70% of what we say, 80% of what we do and say. The suspiciously round numbers are the first clue. Human memory does not arrange itself in multiples of ten.

Researchers who went looking for the source found no study behind them. The percentages appear to have been first published in 1967 by an employee of Mobil Oil Company in the magazine Film and Audio-Visual Communications, presented as established fact without citation.

From there the figures propagated the way misinformation usually does — each new author citing the previous one rather than any original research, until the chain of citations looked like a body of evidence. Similar unsourced percentage claims have been traced in publications going back as far as 1914.

What Dale's Cone Actually Described

Edgar Dale's Cone of Experience is a real and useful model. It arranges learning experiences from the concrete — direct purposeful experience, contrived experiences, dramatised participation — up toward the abstract, ending in visual and verbal symbols.

Dale used the cone's shape to convey the gradual loss of sensory information as an experience becomes more abstract. He described what learning experiences are like, not how much of each one a student retains. He explicitly did not present it as a design tool for ranking instructional methods by effectiveness.

Grafting recall percentages onto that shape reverses its meaning. It turns a descriptive taxonomy into a prescriptive scoreboard, and it implies a precision about memory that no one has ever measured.

What the Meta-Analyses Really Measure

Strip out the myth and the actual research picture is still favourable. Rather than retention percentages, education researchers report effect sizes — a standardised measure of how far the treatment group's outcomes shift relative to a control group.

The numbers are large enough to take seriously, but they vary meaningfully depending on which studies are included and how strictly they are screened. That variation is itself the most useful finding.

Why Effect Size Beats a Retention Percentage

An effect size of 1.07 means the average student in the VR group scored roughly a full standard deviation above the average student in the comparison group. In a normally distributed classroom, that moves a median student to around the 85th percentile. It is a large result by education-research standards, where 0.4 is often treated as the threshold for a practically meaningful intervention.

The reason researchers use this measure instead of a retention percentage is that retention is not a single quantity. It depends on what was learned, how it was tested, and how long after. A number like '90% retained' is meaningless without specifying all three, which is precisely why no credible study reports one.

The trade-off is that effect sizes are harder to put on a poster. That is most of why the fabricated percentages have survived so long — they are memorable, concrete, and completely unencumbered by evidence.

Where VR Clearly Wins — and Where It Doesn't

The aggregate effect sizes hide the single most practical finding in this literature: the benefit is highly dependent on subject matter. VR is not a general-purpose upgrade to teaching. It is a specific tool that solves a specific problem, which is helping students understand things that occupy space.

For spatial subjects the advantage is consistent across multiple studies. For abstract and text-heavy material the advantages are much less clear, and in some comparisons VR performs no better than a well-made diagram at a fraction of the cost.

Why Rigor Cuts the Effect Roughly in Half

The gap between g = 1.07 and SMD = 0.55 is not a contradiction. It is what happens when studies are sorted by experimental quality, and it is a familiar pattern across educational technology research.

Smaller studies with weaker controls tend to report larger effects. They are also more likely to be published when results are positive, which inflates the pooled average. Screening for methodological rigor filters much of that out, and the effect settles at a level that is still clearly worthwhile but distinctly less spectacular.

A moderate effect across 180 primary studies and nearly 19,000 participants is a solid, defensible basis for investment. It is a far better argument than a number someone made up in 1967, and it comes with the honesty that the strongest claims are the least reliable ones.

What a $299 Headset Changed

None of the research mattered much to school budgets while headsets cost as much as a laptop cart. Standalone hardware at around $299 — the launch price of the Meta Quest 3S — moved VR into the same procurement conversation as a Chromebook.

That shift changes the relevant question. Schools no longer have to ask whether VR is transformative enough to justify an exceptional purchase. They have to ask whether, for a particular subject, it beats the alternatives available at similar cost.

For a biology department teaching cellular structure, the evidence supports yes. For a literature department, the same money buys more learning almost anywhere else. Cost parity makes the subject-specificity finding the decisive one rather than an academic footnote.

The Novelty Problem Nobody Measures Well

A persistent weakness in this literature is duration. Most studies measure a single session or a short unit, and headsets are novel to nearly every participant in them. Novelty reliably raises attention and engagement regardless of whether the medium teaches anything better.

Long-term studies of students who use VR routinely enough for it to become ordinary are scarce. Until more of them exist, some portion of the measured effect should be treated as attributable to the newness of the experience rather than to immersion itself.

This is not a reason to dismiss the findings. It is a reason to expect the real steady-state benefit to sit closer to the rigor-adjusted 0.55 than to the headline 1.07 — and to be suspicious of any vendor citing the higher figure without the caveat.

What Actually Replaces the Textbook

The textbook is not dead, and framing it that way misreads the evidence. What the research supports is narrower and more useful: for material with genuine spatial structure, immersion does something a page cannot, and it does so with a measurable, replicated advantage.

For everything else, text remains an efficient, cheap, reviewable, searchable technology that students can navigate at their own pace. No meta-analysis in this literature makes a case for replacing it wholesale, and the ones claiming to usually turn out to be citing 1967.

The realistic 2026 classroom is not a holodeck. It is a room where a headset comes out for the circulatory system and the tectonic plates, and stays in the cupboard for the essay unit — a division of labour the evidence actually supports.

Conclusion

The strongest argument for VR in education has been undermined for years by the weakest possible evidence. A fabricated set of retention percentages, invented in a trade magazine and never sourced to any study, became the industry's favourite talking point precisely because it was easier to remember than the real research.

The real research is better. Large pooled effects across thousands of participants, a moderate effect that survives strict quality screening, and a clear, replicated finding about which subjects actually benefit — that is a more durable case than any invented number, and it is one that holds up when a sceptical administrator asks for the source.

The useful question was never whether immersion beats reading. It is which things are worth immersing in, and for those, the answer is now reasonably well measured.

Data Source and Attribution

Educational Psychology ReviewThe Debunker ClubStanford VHIL

The figures in this story come from peer-reviewed meta-analyses of virtual reality instruction, including a stratified meta-analysis published in Educational Psychology Review integrating 26 first-order meta-analyses, and comparative meta-analytic work published in Applied Sciences. The debunking of the '10% / 90%' retention percentages follows documentation by the Debunker Club and the Learning Development Accelerator, which traced the figures to a 1967 article in Film and Audio-Visual Communications.

FactsFigs reviews, cleans, and cross-checks every source dataset before shaping it into a data story. Each visualization is created and designed in FactsFigs Design Studio — an internal tool developed and owned by FactsFigs — and is the original work of a FactsFigs author, not an AI-generated copy of any existing graphic. Individual assets within a visual may or may not be produced with AI tools, but the design of the visual itself is solely FactsFigs' own.

Figures are estimates at the time of publication, provided for information only — nothing here is financial advice or a guarantee of accuracy.

2026-07-20