Ranking Study Methods by Evidence: What Works and What's Myth (Plain-Language Edition)

This is the condensed edition of the deep dive of the same name. Every key number was independently verified; for full arguments and sources, read the deep dive.

Most study advice you've heard doesn't have the evidence you think it has

Popular culture has standard answers for how to learn: put in ten thousand hours and you'll master anything; everyone has a learning style, and teaching should match it; highlighting and rereading are a student's bread and butter.

Learning science has its own answers — some effects have been measured for over a century — and they barely overlap with the popular ones. This article re-ranks the common study methods by strength of evidence: which ones survive repeated testing, and which were manufactured by bestsellers.

First, three anti-scam questions

Before believing any "method X boosts scores by Y%," ask:

Compared to what? The same "practice quizzing works" scores twice as high against "doing nothing" as against "rereading the textbook" — and its edge nearly vanishes against serious techniques like concept mapping. Marketers always quote the first number.

Measured where? Big effects from lab word-memorization studies often shrink substantially in real classrooms. In field experiments, a method that lifts achievement by "0.1 standard deviations" already beats most educational interventions. Unsexy, but that's the real-world scale.

Who's selling? Bestselling authors, questionnaire vendors, and training companies all report numbers in this field. Even education's most famous "what works" league table (Hattie's Visible Learning) was caught by a statistician computing negative probabilities — and called pseudoscience in print. Rankings can be wrong, spectacularly.

Top of the table: test yourself + space it out

Test yourself (retrieval practice). After reading something, close the book and recall it — far better than reading it again. In the classic experiment, one group read a passage 14 times and remembered 40% a week later; another group read it barely more than 3 times but took three recall tests, and remembered 61%. The kicker: the 14-times group was the most confident — smoothness deceives; effortful recall is what builds memory. Two conditions: the benefit shows up after a delay (same-day, rereading actually wins), and it works best with feedback. This holds across hundreds of real classroom studies, and multiple statistical checks found no "only the good results got published" inflation — rare in education research. And the worry that frequent quizzing causes anxiety? Measured: practice tests reduce test anxiety.

Space it out (spaced practice). Three reviews spread over three days beat three reviews in one sitting. This is one of psychology's oldest, best-documented effects. Two practical calibrations: set review gaps at roughly 20% of the time until you need the material (a month out? review every few days); and the popular app rule that "gaps must keep expanding" turns out to perform no better than evenly spaced reviews. One cold shower: when spacing was casually added to regular university courses, average gains were about 2 percentage points. The method is real; effortless magic is not.

Mix it up (interleaving). When practicing math, mix problem types instead of finishing one type before the next. It owns one of the prettiest results in education research: 787 middle-schoolers, a preregistered randomized trial, four months of practice — on a surprise test a month later, the mixed classes scored 61% versus 38% for the blocked classes. But the borders are sharp: it only helps when the material is confusable and needs telling apart (math problem types, painters' styles); for vocabulary, blocking actually wins. And mixing takes longer and feels worse while you do it — again, "feels bad, works well."

Middle of the table: right direction, inflated numbers

Think in class (active learning). Hundreds of university STEM studies say classes where students solve, discuss, and answer beat pure lecturing by about 0.47 standard deviations on average, with failure rates dropping from 34% to 22%. The direction is credible, with two discounts. First, nine-tenths of those studies weren't strict randomized experiments, so don't treat the numbers as precise. Second, a Harvard experiment with identical handouts found the active-learning group measurably learned more but felt they learned less — and rated the class lower. The cozy "I'm learning so much" feeling in a smooth lecture is largely an illusion. One more thing: the warring camps actually agree on the core — what works is thinking with guidance. Fully unguided "figure it out yourself" teaching is opposed by both sides. Beginners should study worked examples first; the same method flips effectiveness as you gain expertise.

Bottom of the table: two famous myths

The 10,000-hour rule. Not a scientist's claim — writer Malcolm Gladwell packaged it in 2008 from a 30-person violin study, and the original researcher, Anders Ericsson, publicly rejected any magic in "ten thousand." The follow-up accounting is worse: pooling nearly a hundred studies, practice amounts explain only about 14% of the difference in achievement — practice matters a lot, but nowhere near "explains everything." And when the violin study was redone in 2019 with stricter methods, the very best players had not practiced more than the merely good ones. Practice is necessary, not sufficient; talent and practice amplify each other rather than compete.

Learning styles. "Visual learners need diagrams, auditory learners need lectures" — the most widely believed teaching theory on Earth: around nine in ten educators endorse it in surveys across countries. Yet properly designed tests (assess style, randomize teaching method, common exam) have failed for decades; the friendliest analysis ever, in 2024, found only a borderline sliver of an effect, and its own authors concluded it's "too small and too infrequent" to justify adopting. Even the designer of the VARK style questionnaire admitted there's no reliable research showing that knowing your style helps you learn. The theory's real talent isn't helping people learn — it's spreading, mainly through teacher training itself.

Footnote: rereading and highlighting aren't condemned, just the worst value for effort — they're what students use most, yet deliver least reliably. Swapping "read it again" for "close the book and recall" is the highest-return trade in studying.

How to check this article's judgments

Twelve testable claims, hardest evidence first:

1. Self-testing and spacing are the two best-evidenced methods — multiple large syntheses converge, they hold in classrooms, and bias checks came back clean.
2. Effect numbers depend on the comparison: 0.61 versus doing nothing, 0.33 versus rereading, near zero versus serious study methods — always ask "compared to what."
3. Testing's benefit needs a delay — same-day, rereading wins; confidence and results often point opposite ways.
4. "Gaps must keep expanding" has no evidence — it ties with even spacing; the "optimal gap" is just a rough ratio that shifts with your deadline.
5. Lab effects shrink in classrooms, by different amounts: self-testing loses a third, casually added spacing drops to ~2 percentage points, and mixing practice is the rare survivor.
6. Mixing has hard borders: the material must be confusable; for vocabulary, blocking wins — and mixing costs more time.
7. Practice amounts explain only ~14% of achievement differences (the corrected figure), and the redo of the original violin study found the top group hadn't practiced more at all.
8. The amount people practice is itself 40–70% heritable; talent and practice amplify each other rather than compete.
9. Learning-styles matching keeps failing qualified tests while ~90% of educators believe it — spread mainly by teacher training.
10. Active learning: direction solid, numbers soft — nine-tenths of the evidence is loosely controlled; and students who learned more felt they learned less, so course ratings punish good teaching.
11. The real consensus across warring camps: thinking plus guidance works; novices need examples and structure, experts benefit from harder modes.
12. Education research itself has systemic flaws: 0.13% of papers are replications, and the most famous rankings contain arithmetic errors — audit any league table before trusting it.

Worth watching: whether the authoritative 2013 technique ratings get an update; whether "mix it up" lands a second strict classroom trial outside middle-school math; and whether AI learning tools ship self-testing + spacing + feedback as defaults — or keep selling "personalized learning styles."

The things that matter most

1. Swap "read it again" for "close the book and recall it," then check your answers — the hardest-evidence, lowest-cost upgrade available.
2. Spread reviews across the calendar, gaps at roughly a fifth of the time until you need it; don't fuss over expanding intervals.
3. Feeling of effort ≠ learning badly; feeling of ease ≠ learning well — distrust any method promising effortless learning.
4. Don't buy courses or sort kids by "learning style" — that money and time beats any of retrieval, spacing, or worked examples exactly never.
5. Practice hard, but skip the 10,000-hour quota — direction, feedback, structure, and fit matter as much as hours.
6. Novices and experts need different medicine: start with worked examples and guidance; add self-testing, mixing, and difficulty as you level up.