Learning & Education

Students of equal ability think less of their own academic ability in a school where their classmates are stronger.

Marsh and Parker reported it in 1984, and Marsh and Hau later found it in each of 26 countries in an international survey. Academic self-concept is formed by comparison with the students actually in the room, so the same child ranks herself lower among high achievers. Davis had seen the career version in 1966: graduate-school plans tracked a student's rank within his college more than the quality of the college.

In practice

• A department that hires only from the strongest programmes finds its junior staff less confident than equally able hires at a weaker rival.
• A pupil who topped her village school moves to a selective one, finishes mid-table, and decides she is not a maths person.

Attributed to

Herbert Marsh; after James Davis's frog-pond study

Also called

Frog-pond effect

Standing

The effect on self-concept is robust; how much it changes later achievement and careers is debated. Compare the mismatch hypothesis, which makes a related claim about outcomes.

Page for this entry

Pupils tutored one-to-one under mastery conditions were reported to outperform conventionally taught pupils by around two standard deviations.

Bloom framed it as a challenge rather than a finding: tutoring works, and does not scale, so the task is to find group methods that approach it. The framing has outlasted the number, and most citations quote the number.

In practice

• A school buys a tutoring programme on a promise of two standard deviations and measures a third of one.
• A method that works with one graduate student per pupil tells you little about a class of thirty.

Attributed to

Benjamin Bloom

Also called

The 2 sigma problem

Possible explanations

One explanation is pacing and feedback: a tutor adjusts to the pupil continuously, which a class cannot. Another is mastery gating, since the pupil does not advance until the material is learned.

In its favour

Tutoring does work: meta-analyses find substantial average effects for structured one-to-one tutoring, and the challenge Bloom set, to find group methods that approach it, remains a useful one.

Against it

The two-sigma figure came from a small number of short studies run by his own graduate students, and later work generally finds effects well below it, though the number is still quoted as an established result.

Standing

The two-sigma figure comes from a small number of short studies with his own graduate students, and later work has generally found smaller effects. The problem it poses is real; the size is not settled.

Page for this entry

Reducing class size in the earliest years produced measurable attainment gains in a randomised trial.

Project STAR randomised pupils and teachers to small and regular classes across Tennessee from 1985, and found gains for the small classes that persisted for several years. California's later statewide reduction found little, in a setting where the policy itself changed who was teaching.

In practice

• A trial that holds teacher quality constant and a policy that cannot are measuring different things.
• A school that halves class sizes by hiring in a shortage may have traded the effect it was buying.

Attributed to

Tennessee Project STAR

Also called

Project STAR; small class effects

Possible explanations

One explanation is attention per pupil, which rises as numbers fall. Another is behaviour, since a smaller room is easier to manage. A third is teacher supply, which explains why the result weakens at scale: a state that halves class sizes must hire many more teachers, and the additional ones are less experienced.

In its favour

Project STAR randomised pupils and teachers, which few education studies manage, and found gains in the earliest years that persisted for several years afterwards.

Against it

Statewide reductions elsewhere produced much smaller or no gains, the cost per point of attainment is high compared with other interventions, and the teacher-supply constraint means the trial result may not transfer to policy.

Standing

The STAR result is among the strongest evidence in education and has not generalised cleanly: statewide reductions elsewhere produced much smaller effects, partly because they forced hiring of less experienced teachers.

Page for this entry

Differences in pupils' family background explain much more of the variation in their achievement than differences in school resources.

Equality of Educational Opportunity, the 1966 survey Congress commissioned under the Civil Rights Act and James Coleman led, covered hundreds of thousands of pupils and was expected to document unequal school funding. Its central finding was that spending, facilities and curriculum explained little of the variation in achievement once family background was accounted for, and that the social make-up of a pupil's classmates mattered more.

In practice

• A district spends a bond issue on new buildings and is surprised when results in the new schools track the old ones.
• Two siblings at different schools end up with grades much closer to each other than to their classmates'.

Attributed to

James S. Coleman and colleagues

Also called

Equality of Educational Opportunity

Possible explanations

One explanation is that the resources the survey measured, such as buildings and spending, are weak proxies for what schools actually do. Another is that family background supplies the preparation, expectation and support that schooling builds on.

In its favour

It was an unusually large survey for its time, its central finding has been reproduced in later work on the limits of pure resource measures, and it redirected education research towards what happens in classrooms.

Against it

Later work found teachers differ greatly in effectiveness, which its measures could not capture, so the conclusion that schools matter little goes beyond what the data showed.

Standing

Later work found that teachers differ greatly in effectiveness, which the report's measures of resources did not capture, so schools matter more than its headline suggests. Compare social capital, Coleman's later concept.

Page for this entry

Students admitted to highly selective colleges who attend less selective ones later earn about as much as those who attend, except students from disadvantaged backgrounds.

Dale and Krueger compared students admitted to the same colleges who chose differently, in a 2002 paper and a 2014 follow-up using administrative earnings records. For most, the more selective college added nothing measurable to average earnings: the ability and ambition that won admission did the work. For Black and Hispanic students and the children of less-educated parents, the selective college did raise earnings.

In practice

• A family remortgages for the more prestigious offer when their daughter's own record was what employers were going to pay for.
• Two friends admitted to the same elite university make different choices, and ten years later earn much the same.

Attributed to

Stacy Dale & Alan Krueger

Also called

The selectivity premium is selection

Possible explanations

One explanation is selection: the ambition and ability that win admission produce the later earnings, not the college. Another is that employers can observe those qualities directly, so the credential adds little for candidates who have alternatives.

In its favour

The design is unusually strong, comparing students admitted to the same colleges who chose differently, and the 2014 follow-up used administrative earnings records rather than self-reports.

Against it

Chetty, Deming and Friedman found in 2023 that elite colleges markedly raise the chance of reaching the very top of the distribution even where average earnings are unaffected, so the result holds at the average and not in the tail.

Standing

Chetty, Deming and Friedman found in 2023 that Ivy-Plus colleges have little effect on average earnings but make students markedly more likely to reach the top 1 per cent of earners, attend an elite graduate school and work at a prestigious firm, outcomes Dale and Krueger's data could not see. The finding holds for the average and not at the top.

Page for this entry

Conditions that slow learning in the session often improve retention and transfer afterwards.

Bjork's distinction is between performance during practice and learning measured later, and the two frequently move in opposite directions. Spacing, interleaving, testing and reduced feedback all make the session feel worse and the retention better, which is why learners reliably choose the method that teaches them least.

In practice

• A trainee who rereads feels fluent and recalls less than one who closed the book and struggled.
• A course that gets the best feedback forms produces the worst results at six months.

Attributed to

Robert Bjork

Also called

Productive difficulty

Further reading

Wikipedia

Page for this entry

Teaching a subject deepens the teacher's own grasp of it, and even preparing to teach improves learning.

Seneca wrote to Lucilius that people learn while they teach; the two-word Latin tag is the later compression. Chase and colleagues found in 2009 that pupils worked harder learning material for a computer character they had to teach than for themselves, which they called the protégé effect. Nestojko and colleagues found in 2014 that students who expected to teach a passage recalled more of it, and organised it better, than students who expected a test.

In practice

• A firm that has each intake run the induction session for the next finds the teachers remember the procedures best.
• A student explaining an exam topic to a friend finds the gaps in her own understanding halfway through the explanation.

Attributed to

Seneca; tested as the protégé effect

Also called

By teaching, we learn; the protégé effect; learning by teaching

Standing

Much of the gain comes from preparing to explain and from retrieval rather than from the act of teaching itself. Compare the illusion of explanatory depth, which explaining exposes.

Page for this entry

An audience rates a lecturer highly for an expressive, confident delivery even when the lecture has no real content.

In 1973 Naftulin, Ware and Donnelly hired an actor, billed him as Dr Myron L. Fox, and had him lecture on mathematical game theory as applied to physician education, in a talk built from double-talk, invented words and contradiction. Audiences of psychiatrists, psychologists and educators rated it favourably. Later studies of what Ware and Williams called educational seduction found that expressiveness raises ratings more than it raises learning.

In practice

• A university that promotes on student evaluations rewards the lecturer with the best jokes, whether or not her students pass the next course.
• A conference audience leaves a keynote enthused and, asked the next morning, cannot say what was claimed.

Attributed to

Naftulin, Ware & Donnelly

Also called

Educational seduction

Standing

A demonstration with small audiences and no control group. The durable finding is the narrower one: delivery moves ratings more than it moves learning. Compare the seductive details effect.

Further reading

Wikipedia

Page for this entry

Instructional support that helps a novice hinders an expert, so the method must change as competence grows.

The scaffolding that carries a beginner becomes redundant material the expert must process and discard, which costs more than it gives. It is the reason a single best teaching method cannot exist: the correct method is a function of where the learner already is.

In practice

• A checklist that halves errors for new staff slows the experienced ones and they stop using it.
• A manual written for beginners is unusable by the people who need the reference.

Attributed to

Slava Kalyuga and colleagues

Also called

The expertise reversal

Further reading

Wikipedia

Page for this entry

Test-score gains from early-childhood programmes largely disappear within a few years, even where some later-life benefits persist.

The randomised Head Start Impact Study found in 2010 that children's early cognitive advantages had largely gone by the end of first grade. Deming's 2009 comparison of siblings nonetheless found that Head Start participants did better as young adults, which is the puzzle: effects that vanish from tests and reappear in outcomes. Tennessee's statewide pre-kindergarten trial went the other way, with participants scoring below the control group by sixth grade.

In practice

• A city funds a pre-school programme on its early test gains and has to decide what to do when the gains are gone by fifth grade.
• A child drilled in reading before school starts reads early, and by eight the classmates who started later have caught up.

Attributed to

Early-childhood evaluation research

Also called

Head Start fade-out; fadeout; convergence

Possible explanations

Candidate explanations are that the early gains were in skills schooling later teaches everyone, that control-group children catch up, that the tests measure what fades rather than what lasts, and that benefits move from test scores into behaviour and later outcomes.

In its favour

The pattern is documented in randomised evaluations, including the Head Start Impact Study, and the later-life gains found in sibling comparisons are real.

Against it

Which explanation holds is unresolved, and Tennessee's statewide pre-kindergarten trial found participants below the control group by sixth grade, which none of the reassuring accounts predicts.

Standing

Whether fade-out means the gains were never real, that later schooling lets the others catch up, or that benefits move from test scores into behaviour and later outcomes is the live question. Compare the Heckman curve.

Page for this entry

Material a learner produces is remembered better than the same material read.

Slamecka and Graf's 1978 experiments had subjects either read a word pair or generate the second word from a rule and a fragment. The generated items were remembered better, and the gap held across variations. The cost is time, which is why generation is dropped first when a syllabus is crowded.

In practice

• A revision session spent producing questions beats one spent reading the answers.
• A team that writes its own checklist follows it; the same checklist handed down is filed.

Attributed to

Norman Slamecka & Peter Graf

Also called

Production effect; self-generation

Further reading

Wikipedia

Page for this entry

Low expectations held by a supervisor depress the performance of those supervised.

Babad, Inbar and Rosenthal's 1982 study of gym teachers found that teachers holding low expectations taught and assessed those pupils differently. Expectancy effects are context-dependent, and there is no direct evidence for ranking this against the Pygmalion effect.

In practice

• A supervisor who has decided a trainee will not make it assigns them less and reviews them less carefully, and is proved right.
• A teacher told a class is a low set covers less material, and the class ends the year having covered less.

Attributed to

Babad, Inbar & Rosenthal

Standing

The negative counterpart of the Pygmalion effect; expectancy effects are context-dependent and not large enough to rank the two without direct evidence.

Further reading

Wikipedia

Page for this entry

Errors made with high confidence are corrected more reliably than errors made with low confidence, once the right answer is supplied.

The result inverts the intuition that strongly held mistakes are the hardest to shift. Butterfield and Metcalfe found that being wrong while certain produces surprise, and surprise captures attention and improves encoding of the correction. The practical implication for teaching is that a confident wrong answer is a better learning opportunity than a hesitant one.

In practice

• An engineer who was sure about the failure mode and wrong remembers the real cause for years.
• A pupil who shrugs at being corrected retains less than one who was certain and is embarrassed.

Attributed to

Brady Butterfield & Janet Metcalfe

Also called

High-confidence error correction

Page for this entry

People believe they understand how everyday things work in far more detail than they can explain.

Rozenblit and Keil asked people in 2002 to rate their understanding of zippers, flush toilets and cylinder locks, then to write out step by step how each works. The ratings fell sharply once the explanation had to be produced. The illusion was specific to explanations of how things work: people were far better calibrated about facts, procedures and stories.

In practice

• A committee confident that it understands its pension scheme cannot agree, when asked, on what happens to a member who changes employer.
• A student who nodded through a lecture on how vaccines work discovers, explaining it to a younger brother, that the lecture felt clearer than it was.

Attributed to

Leonid Rozenblit & Frank Keil

Also called

IOED; the knowledge illusion

Standing

Compare the illusion of fluency, its study-room cousin: both mistake familiarity for understanding. Several entries lean on it, including cached thoughts.

Page for this entry

Ease of processing is mistaken for depth of learning, so learners choose the methods that feel best and teach least.

Rereading and highlighting produce a strong sense of familiarity and poor retrieval; testing and spacing produce the reverse. Because learners judge their learning by how the session felt, the sensation is systematically anticorrelated with the outcome, which is why revision advice is so hard to follow.

In practice

• A student who has read the chapter four times cannot answer a question about it and is genuinely surprised.
• A team that attends a fluent presentation rates it highly and cannot reconstruct it a week later.

Attributed to

Metacognition research; Bjork and colleagues

Also called

Fluency illusion; judgement of learning error

Standing

Compare the illusion of explanatory depth: both mistake familiarity for understanding.

Page for this entry

Mixing problem types within a practice session produces worse practice performance and better later retention than practising one type at a time.

Blocked practice lets the learner apply a method without having to select it, which is the part the test will require. Interleaving forces selection every time. The effect is consistent enough to be a teaching recommendation and unpopular enough that few adopt it, because it looks like failure while it is happening.

In practice

• A maths class that mixes topics scores worse on the Friday quiz and better on the summer paper.
• A trainee drilled on one fault type diagnoses it perfectly and cannot tell which fault they are looking at.

Attributed to

Cognitive psychology of learning; Rohrer and Taylor among others

Also called

Mixed practice; varied practice

Page for this entry

Learning occurs best when a person is mentally prepared.

Thorndike's third law, and the one that aged worst — it is closer to a statement about motivation than a testable regularity, and modern instructional design uses prerequisite knowledge instead.

In practice

• A trainee sent on the advanced course without the introductory one sits through three days and retains the coffee breaks.
• A class taught long division before it is secure with subtraction learns a procedure it has no way to check.

Attributed to

Edward Thorndike

Page for this entry

A young child's ability to wait for a larger reward is claimed to predict achievement and success years later.

Mischel's experiments at Stanford's nursery school offered a child one treat now or two for waiting. Follow-ups in 1988 and 1990 linked longer waiting to higher SAT scores in adolescence, and the test became shorthand for self-control as destiny. In 2018 Watts, Duncan and Quan repeated it with a far larger and more varied sample: the association was half the original size, and was cut by two-thirds once family background, early cognitive ability and the home environment were controlled.

In practice

• A school buys a self-control curriculum on the strength of the marshmallow studies, when much of what the test measured was the children's circumstances.
• A parent worries that a four-year-old who ate the biscuit at once is headed for trouble, when the more reliable signal in the test was everything else about the family.

Attributed to

Walter Mischel and colleagues

Also called

Stanford marshmallow experiment; delay of gratification

Possible explanations

One explanation is self-control as a stable trait. Another is circumstance: a child who has learned that promised treats do not arrive has good reason not to wait, which experiments using unreliable adults support. A third is that waiting and later achievement both follow from family resources.

In its favour

The original studies were carefully run, and the follow-ups did find an association with later achievement.

Against it

The 2018 replication found that association about half the size, and a third of that again once family background and early ability were controlled, so the test measures circumstances as much as willpower.

Standing

The 2018 replication found a small remaining association, concentrated in whether a child could wait at least twenty seconds at all. The story of willpower as destiny rested largely on the original sample's advantages.

Further reading

Wikipedia

Page for this entry

Early reading skill compounds, because children who read well read more and thereby widen the gap.

Stanovich's 1986 paper applied Merton's Matthew effect to literacy: the child who decodes fluently reads for pleasure, acquires vocabulary and background knowledge, and pulls away, while the struggling reader avoids the activity that would close the gap. The mechanism is plausible and widely accepted; the divergence it predicts is not consistently observed.

In practice

• A child who reads a million words a year and one who reads twenty thousand are not receiving the same curriculum.
• An intervention at seven costs a fraction of the same intervention at fourteen, and the difference is compounding.

Attributed to

Keith Stanovich

Also called

Rich-get-richer in literacy

Possible explanations

The explanation is compounding through exposure: a child who decodes fluently reads more, which builds vocabulary and background knowledge, which makes the next text easier.

In its favour

The mechanism is well specified and the early gap in decoding is real and measurable, which is why early reading screening is now standard.

Against it

Evidence for widening gaps is mixed, since some studies find divergence and others stable or narrowing differences, and the result depends heavily on which test and which scale are used.

Standing

Evidence for widening gaps is mixed: some studies find divergence, others find stable or narrowing differences, and the result depends heavily on the measure and the period.

Page for this entry

Struggling with a problem before being taught how to solve it produces deeper conceptual understanding than being taught first, provided the teaching follows.

Kapur introduced it in 2008 with Singapore pupils who attempted complex problems before instruction: they did worse during the lesson and better on the later tests of understanding. Sinha and Kapur's 2021 meta-analysis, pooling 166 comparisons from 53 studies, found an advantage for conceptual knowledge and transfer, largest where the design followed the method closely.

In practice

• A statistics teacher asks the class to invent a measure of how spread out some scores are before showing them the standard deviation, and the formula arrives as the answer to a question they already have.
• Someone who spent an evening failing to fix a dripping tap understands the plumber's explanation the next day in a way they would not have cold.

Attributed to

Manu Kapur

Also called

Problem solving before instruction

Possible explanations

The explanation is preparation for learning: attempting a problem first activates prior knowledge and exposes the gaps, so the instruction that follows lands on a structure ready to receive it.

In its favour

The meta-analysis of 166 comparisons finds an advantage for conceptual knowledge and transfer, largest where the design follows the method closely.

Against it

Cognitive load researchers find explicit instruction first better for novices, the advantage does not appear in procedural skill, and failure without the teaching phase that follows is simply failure.

Standing

Disputed by cognitive load researchers, who find explicit instruction first better for novices. The advantage appears in conceptual understanding rather than procedural skill, and failure without the teaching phase is just failure. Set against the worked example effect.

Page for this entry

The largest educational experiment run found explicit, structured instruction outperforming discovery-based models on basic skills and, unexpectedly, on self-esteem.

Follow Through compared instructional models across sites over years at a scale never repeated. Direct Instruction came out ahead on the measured outcomes including affective ones. The result was disputed methodologically and largely not adopted, so the episode is cited as often for what happened to the evidence as for the evidence.

In practice

• A finding that contradicts a profession's prevailing theory is examined far more closely than one that confirms it.
• An evaluation whose result nobody wanted is answered with methodological objections that were available before the result.

Attributed to

US federal education evaluation

Also called

Direct Instruction evaluation

Possible explanations

The explanation offered by its proponents is that basic skills are best built by explicit sequenced instruction with immediate correction, and that discovery methods leave gaps that compound.

In its favour

It remains the largest educational experiment ever run, it measured affective as well as academic outcomes, and the Direct Instruction result has been replicated in later trials of structured programmes.

Against it

The evaluation has been contested on design, model fidelity and site selection, sponsors implemented their models unevenly, and its findings had almost no effect on practice, which is itself a finding about how evidence is used in education.

Standing

The evaluation has been contested on design, model fidelity and site selection, and its findings had little effect on practice, which is itself the more interesting result.

Page for this entry

Higher expectations lead to increased performance.

Rosenthal and Jacobson's 1968 classroom study is methodologically weak — the test was unpublished and the effect concentrated in the youngest grades. Expectancy effects are real and smaller than the original claim.

In practice

• A teacher told a class is a strong set gives harder work, marks it more closely, and the class outperforms the one taught next door.
• The same teacher given an accurate description of a middling set produces a gap far smaller than the staffroom expects, which is roughly where the honest estimate sits.

Attributed to

Rosenthal & Jacobson

Also called

Rosenthal effect; named from Shaw's play, after Ovid

Possible explanations

One explanation is differential treatment: teachers give pupils they expect more of more time, harder material and warmer feedback. Another is the pupil's own expectation, altered by how they are treated.

In its favour

Expectancy effects are real and do replicate, particularly where the expectation is induced before a teacher has other information about a pupil.

Against it

The original study's test was unpublished and its effect concentrated in the youngest classes, meta-analysis finds small average effects, and teachers' expectations are usually accurate rather than self-fulfilling.

Standing

The original classroom study is methodologically weak; expectancy effects are real but smaller.

Further reading

Wikipedia

Page for this entry

Adding interesting but irrelevant material to a lesson reduces how much of the relevant material is learned.

Garner and colleagues named it in 1989. Harp and Mayer's 1998 experiments added vivid anecdotes about people struck by lightning to a lesson on how lightning forms, and students remembered the anecdotes and understood less of the mechanism. Sundararajan and Adesope's 2020 meta-analysis found a small-to-moderate negative effect across dozens of studies.

In practice

• A training deck opens each section with a funny stock photograph, and the feedback praises the photographs while the quiz scores fall.
• A museum label leads with the curator's anecdote about the discovery, and visitors leave knowing the anecdote and not what the object was for.

Attributed to

Ruth Garner and colleagues; Harp & Mayer

Also called

Seductive details

Standing

Most of the studies used short sessions, so whether the interest such details add pays for itself over a whole course is less tested. Compare the Dr. Fox effect.

Further reading

Wikipedia

Page for this entry

Private tutoring that mirrors the school curriculum grows with the stakes of examinations, and widens the advantage of families who can pay for it.

Stevenson and Baker named it in 1992 from Japan, where the juku cram schools shadowed the school system and shaped who reached university. Bray's 1999 study for UNESCO found the same parallel system across Asia and beyond, largest where a single examination decides a child's future. China banned for-profit tutoring in core school subjects in 2021 in an attempt to contain it.

In practice

• A country raises the stakes of its university entrance exam and watches enrolment at evening cram schools climb in step.
• Parents who swore they would never hire a tutor hire one in the year before the exam, because every other family in the class has.

Attributed to

David Stevenson & David Baker; Mark Bray

Also called

Private supplementary tutoring; juku; hagwon

Standing

Whether tutoring raises achievement by much is disputed; the growth with exam stakes and the link to family income are the robust parts. Compare gyoyukyeol and kiasu.

Page for this entry

Earnings jump in the year a qualification is completed, by more than the extra year of schooling alone would explain.

Hungerford and Solon found in 1987 that American returns to schooling rose sharply at the years that complete a qualification, the twelfth and sixteenth, rather than smoothly year by year. Jaeger and Page confirmed it in 1996 with data on the degrees people actually held. The name comes from the parchment that diplomas were once printed on.

In practice

• A firm pays a graduate markedly more than a candidate who completed all but one module of the same degree.
• A student two courses short of a degree finds that employers treat him as having none.

Attributed to

Thomas Hungerford & Gary Solon

Also called

Diploma effect

Standing

The jump is well established. Whether it shows that diplomas mostly signal ability, or that people who finish differ from those who do not in ways that also raise pay, is disputed. Compare signalling.

Page for this entry

Learning spread over time is retained far better than the same total time massed together.

Ebbinghaus documented it in 1885 and it remains among the most robust findings in learning research. Spaced repetition software is its direct industrial application.

In practice

• A language app that shows a word again after a day, then three days, then a week, is doing the one thing that reliably works.
• A trainee who studies forty minutes daily outperforms one who studies six hours on Sunday, and the second feels more diligent.

Attributed to

Hermann Ebbinghaus

Further reading

Wikipedia

Page for this entry

People given more time for a task start it later, so time added as a safety margin is used up before the work begins.

Goldratt named it in Critical Chain (1997), a business novel about project management, after the student who, given more time for an assignment, still starts it at the last minute. His point was about estimates: padding each task does not protect the schedule, because the padding is spent at the front. The nearest controlled test, Bryan and Locke's 1967 study of Parkinson's law, found that generous time limits slowed people down by lowering the goals they set.

In practice

• A team given four weeks for a feature that needs two starts in the third week, and the buffer its manager added is gone before anyone opens the code.
• A student granted a week's extension on an essay begins it the night before the new deadline, exactly as she would have begun the original one.

Attributed to

Eliyahu Goldratt

Standing

Named from practitioner observation rather than measured. The procrastination literature supports the pattern; whether extra time buys any extra quality is less clear. Compare Parkinson's law, which spreads the work across the time rather than postponing it.

Page for this entry

Pupils lose academic ground over the long summer break, and the loss is claimed to be larger for poorer children, widening gaps.

Heyns documented it in 1978, and Cooper's 1996 meta-analysis put the average loss at about a month of schooling. The strongest version came from Baltimore's Beginning School Study, where the reading gap between richer and poorer pupils more than tripled by eighth grade, with all of the growth appearing to happen over summers. Von Hippel and Hamrock showed in 2019 that much of that pattern came from the way older tests were scaled.

In practice

• A district funds summer school to close its income gap, and finds that the gap in its own spring and autumn scores barely moved over the break.
• A child comes back in September rusty on long division, which is real, and a few weeks of practice restore it.

Attributed to

Barbara Heyns; Harris Cooper and colleagues

Also called

Summer slide; summer setback

Possible explanations

Candidate explanations are simple forgetting over a long break, differences in what children do during the summer, and measurement, since tests scaled in older ways can manufacture the appearance of loss.

In its favour

Some loss appears on some tests, and the mechanism is plausible for skills that need practice, such as arithmetic.

Against it

Findings often fail to replicate in recent data, much of the original pattern came from test scaling now considered artefactual, and whether summers widen gaps by income is unsettled.

Standing

Workman, von Hippel and Merry found in 2023 that findings on summer loss often fail to replicate even in recent data. Some loss appears on some tests; whether summers widen gaps by family income is unsettled.

Further reading

Wikipedia

Page for this entry

A substantial share of students admitted to college who intend to enrol in the autumn never arrive, lost over the summer to paperwork, costs and the absence of support.

Castleman and Page, in Summer Melt (2014), put the loss at between a tenth and two-fifths of college-intending school leavers in the districts they studied, concentrated among low-income students. The summer is the one stretch when a school counsellor is no longer responsible for a student and a college is not yet. Their text-message reminders raised enrolment modestly in several districts.

In practice

• A high school counts its leavers' college acceptances in June and never learns how many enrolled in September.
• A first-generation student misses a housing deposit deadline in July, takes a job instead, and tells herself she will start next year.

Attributed to

Benjamin Castleman & Lindsay Page

Standing

Reminder campaigns that helped in pilots have shown little effect when run at national scale, which suggests reminders are not the whole problem.

Page for this entry

Retrieving information strengthens memory more than re-reading it does.

Roediger and Karpicke's 2006 finding: students who read once and tested repeatedly outperformed those who read repeatedly, despite feeling less confident. The feeling of fluency is what misleads.

In practice

• A trainee who closes the manual and writes down what they remember retains more than one who reads it four times.
• The feeling of fluency is what misleads: re-reading feels like learning and retrieval feels like failure.

Attributed to

Roediger & Karpicke

Further reading

Wikipedia

Page for this entry

Retention is claimed to rise from a tenth of what is read to nine-tenths of what is taught to others.

Dale's Cone of Experience, to which the pyramid is routinely credited, contained no percentages in its 1946, 1954 or 1969 versions, and Dale stated it was not a ranking of effectiveness. Letrud traces the pyramid to a synthesis of the cone with a separate and equally unsourced retention chart. The numbers are all multiples of five and sum with a neatness real retention data never shows.

In practice

• A training budget is allocated on the strength of a diagram whose owner has said it cannot produce the study behind it.
• A department drops lectures for group work citing 5% against 50%, and the two figures were never measured.

Attributed to

Attributed to the NTL Institute; wrongly credited to Edgar Dale

Also called

Cone of learning; the retention pyramid

Possible explanations

One explanation for its spread is that the ranking matches teachers' intuitions and gives a simple rule. Another is authority, since the numbers travelled attached to a respected institute's name.

In its favour

The general point, that active practice and explaining to others beat passive reading, is supported by research on retrieval practice and on learning by teaching.

Against it

The figures have no traceable source, the institute could not produce the research when asked, and the cone of experience it is credited to contained no percentages and was never a ranking of effectiveness.

Standing

The figures have no traceable source. NTL, asked for the study, said it believed the numbers accurate but could no longer find the research, and later dropped the pyramid for having no scientific foundation.

Further reading

Wikipedia

Page for this entry

Listening to Mozart was reported to raise spatial-reasoning scores briefly.

The 1993 report described a temporary improvement on a paper-folding task lasting minutes. It became state policy in places — free classical recordings for newborns — on the strength of a claim the paper did not make. Later work suggests any effect is arousal and mood rather than Mozart.

In practice

• A nursery plays classical music at nap time and reports it as an intervention.
• A finding about one task in adults for ten minutes is legislated for as a claim about children's development.

Attributed to

Rauscher, Shaw & Ky

Also called

Music and spatial reasoning

Possible explanations

The best-supported explanation is arousal and mood: any enjoyable stimulus that raises alertness briefly improves performance on some tasks, and the music is incidental.

In its favour

The original experiment was real and carefully reported, and the arousal mechanism it turned out to demonstrate is itself a genuine finding.

Against it

The effect was small, lasted minutes and applied to one spatial task; meta-analysis finds little support, and the popular claim that it raises intelligence in children has no basis.

Standing

The original effect was small, short-lived and confined to one spatial task; meta-analysis found little support, and the popular claim that it raises intelligence in children was never made by the original study.

Further reading

Wikipedia

Page for this entry

Prior learning influences the ability to learn new tasks.

Thorndike's finding, and his identical-elements theory limited it sharply: transfer occurs where the tasks share components, which is why "learning to think" through Latin did not deliver what was claimed.

In practice

• A firm that teaches its analysts chess to sharpen their thinking gets better chess players.
• Learning Latin does not make you think better, and learning one programming language does help with the next, because the second shares components and the first does not.

Attributed to

Edward Thorndike

Standing

The finding that retired the doctrine of formal discipline, which held that hard subjects such as Latin train the mind in general.

Further reading

Wikipedia

Page for this entry

Many academically strong students from low-income families enrol at colleges well below the selectivity they could gain admission to, or apply to none that would stretch them.

Bowen, Chingos and McPherson named it in Crossing the Finish Line (2009): in North Carolina, most qualified students from the lowest income quarter undermatched, against about a quarter from the highest, and undermatched students graduated less often. Hoxby and Avery showed in 2012 that most high-achieving, low-income students never apply to a selective college, often because none recruits where they live, and Hoxby and Turner found that posting them application guidance and fee waivers changed where they applied.

In practice

• A rural valedictorian applies only to the regional college her teachers attended, and never learns that the flagship would have cost her less after aid.
• A top student chooses the college his friends are going to because nobody in his family has been through an admissions process.

Attributed to

William Bowen, Matthew Chingos & Michael McPherson

Also called

College undermatch; the missing one-offs

Possible explanations

One explanation is information: students who know nobody who attended a selective college underestimate both their chances and the aid available. Another is cost, since the sticker price is visible and the discount is not. A third is preference for staying close to home.

In its favour

The pattern is documented in state administrative data, and mailing application guidance and fee waivers changed where students applied, which shows the barrier is partly informational.

Against it

Bastedo and Flaster argued in 2014 that the concept rests on contestable assumptions about what counts as a match, and whether undermatching costs students anything depends on the Dale-Krueger question, where the premium is largest for exactly this group.

Standing

Bastedo and Flaster argued in 2014 that the concept rests on contestable assumptions about what a match is. Whether undermatching costs students anything turns partly on the Dale–Krueger finding, where the selective-college premium is largest for exactly these students.

Further reading

Wikipedia

Page for this entry

Novices learn more from studying fully worked solutions than from attempting equivalent problems unaided.

A robust result within cognitive load theory: the novice attempting a problem spends capacity searching rather than learning the structure, while the worked example makes the structure visible. It reverses with expertise, which is the expertise reversal effect.

In practice

• A class given six solved examples outperforms the class given six problems and the same time.
• An apprentice shown the finished joint and how it was cut learns faster than one told to work it out.

Attributed to

John Sweller and colleagues

Also called

Worked examples over problem solving

Standing

Set against productive failure, which claims that struggling first works better for conceptual understanding.

Page for this entry

An examination format rigid enough to be graded consistently comes to be mastered for its own sake, and crowds out the thinking the examination was meant to find.

The eight-legged essay was the required form of answer in the Ming and Qing civil service examinations from the late fifteenth century: an exposition of a Confucian text in eight fixed sections with prescribed parallel phrasing. It made the grading of enormous numbers of candidates tractable and comparable, and it was learned as a template. The seventeenth-century scholar Gu Yanwu judged its damage greater than the burning of the books; it was abolished in 1901.

In practice

• An exam board's mark scheme rewards a fixed essay structure so reliably that schools teach the structure, and the arguments inside it become interchangeable.
• A student who has learned the five-paragraph essay writes one in response to every prompt, including a request for a letter.

Attributed to

Ming and Qing civil examinations

Also called

Eight-legged essay; the examination essay

Standing

Defenders then and since point to what it solved: grading hundreds of thousands of candidates on a common standard. Its rigidity was the price of the examination's fairness. Compare Campbell's law.

Further reading

Wikipedia

Page for this entry

Increases in federal student aid enable colleges to raise their prices, so part of the aid is captured by the institution rather than the student.

William Bennett, then Secretary of Education, made the argument in a 1987 New York Times article, 'Our Greedy Colleges'. The evidence is strongest where colleges depend most on aid: Cellini and Goldin found in 2014 that aid-eligible for-profit colleges charged more than comparable ineligible ones by about the value of the aid, and Lucca, Nadauld and Shen estimated that each dollar of expanded subsidised loans raised list prices by about sixty cents.

In practice

• A for-profit college raises its fees by roughly the size of a grant increase in the year the grant rises.
• A family's larger loan allowance is matched by a larger bill, and the student is no better off.

Attributed to

William Bennett

Also called

Aid-driven tuition

Possible explanations

The explanation is capture: aid that follows the student to whichever institution they choose raises what students can pay, and institutions with pricing power can absorb part of it.

In its favour

The evidence is strongest where colleges depend most on aid, with studies finding aid-eligible for-profit institutions charging more than comparable ineligible ones, and estimates of substantial pass-through into list prices.

Against it

Studies of public and non-profit colleges often find little effect, and list prices can rise while discounted net prices do not, so the hypothesis holds where aid is a large share of revenue rather than generally.

Standing

Studies of public and non-profit colleges often find little effect, and list prices can rise while discounted net prices do not. The hypothesis holds best where aid is a large share of a college's revenue.

Page for this entry

A ranking of educational objectives from recall through to evaluation, used to write and grade assessments.

Produced in 1956 by a committee of examiners looking for a common vocabulary for exam questions. Its durability comes from solving a practical problem — two examiners can now agree what a question is asking — rather than from evidence that minds work in that order.

In practice

• A paper that tests only recall is visible as such the moment the questions are classified.
• A course claiming to teach evaluation sets an exam answerable from the glossary.

Attributed to

Benjamin Bloom and colleagues

Also called

The cognitive domain taxonomy

Standing

The hierarchy is a classification adopted by committee rather than a measured developmental sequence, and the assumption that lower levels must precede higher ones is not established.

Further reading

Wikipedia

Page for this entry

A university's costs are set by its revenues: it raises all the money it can and spends all it raises.

Howard Bowen set it out in The Costs of Higher Education (1980) as five laws: the dominant goals are excellence, prestige and influence; there is no limit to the money that could be spent pursuing them; each institution raises all it can; each spends all it raises; and the cumulative effect is ever-rising expenditure. Because quality has no agreed measure, no sum is ever enough.

In practice

• A university receives a large bequest and within a few years has spent it on new programmes and buildings without lowering its fees.
• A department whose budget is cut finds savings it said did not exist, and when the budget is restored the savings vanish.

Attributed to

Howard Bowen

Also called

Bowen's law; the revenue theory of cost

Standing

The rival explanation for rising university costs is Baumol's cost disease, which puts it down to labour-intensive teaching rather than appetite. Compare Parkinson's second law, of which this is the institutional case.

Page for this entry

Working memory is the bottleneck in learning, so instruction must strip extraneous load.

Sweller's finding that worked examples beat problem-solving for novices, because the search for a solution consumes the capacity needed to learn the method. It inverts a good deal of discovery-learning practice.

In practice

• A training course opens with a live exercise before teaching the method, and the trainees spend the hour searching rather than learning.
• An onboarding document explains the architecture, the deployment process and the on-call rota on day one, and none of it is retained.

Attributed to

John Sweller

Further reading

Wikipedia

Page for this entry

Middle-class child-rearing that schedules activities and coaches children to question professionals transmits an ease with institutions that schooling then rewards.

Lareau's Unequal Childhoods in 2003 contrasted concerted cultivation with what she called the accomplishment of natural growth, and found the advantage did not lie in the activities but in the practised habit of addressing adults in authority as equals and negotiating on one's own behalf. The children who had it extracted more from every institution they met.

In practice

• Two equally able pupils get different outcomes from the same appeal process, and the difference is which one has seen an adult argue with an official.
• A patient who asks the consultant three follow-up questions receives a different consultation from one who does not.

Attributed to

Annette Lareau

Also called

Class parenting styles; accomplishment of natural growth

Possible explanations

One explanation is practice: children coached to question professionals arrive at institutions already able to do it. Another is scheduling, which teaches negotiation with adults and the management of time.

In its favour

Later survey work finds class differences in parenting practice, and in children's ease with institutions, broadly as described, and the mechanism is specific enough to test.

Against it

The study observed twelve families, and the link from practice to outcome is inferred rather than measured. Parenting style is also hard to separate from income, neighbourhood and inherited advantage.

Standing

Drawn from intensive observation of a small number of families; the ethnographic depth is the strength and the sample is the limit.

Further reading

Wikipedia

Page for this entry

Open-access colleges manage the gap between students' ambitions and their chances by steering the unsuccessful gradually towards lesser goals rather than failing them outright.

Clark described it in 1960 from a study of a California junior college, borrowing the phrase from Goffman's account of how confidence men calm a victim so that he goes away quietly. Counselling, testing and probation redirected students on the transfer track towards vocational programmes, so that failure was experienced as a choice. The college kept its promise of access while rationing the degrees it led to.

In practice

• A student told each term that a certificate might suit her better than transfer takes the certificate, and remembers deciding.
• A company that never dismisses anyone offers its weakest performers a series of sideways moves until they resign.

Attributed to

Burton Clark

Also called

Cooling the mark out

Standing

Later research also finds colleges that raise students' ambitions, and disputes how deliberate the steering is. Compare maximally maintained inequality.

Page for this entry

Expanding education raises the qualification demanded for a given job faster than the job's content changes, so credentials rise without skills doing so.

Collins argued in 1979 that credentials function largely as a screening and closure device rather than as evidence of trained capability. As supply grows, the threshold rises to preserve the signal, so each generation must go further to reach the same position. It predicts the observed decoupling between years of schooling and job content.

In practice

• A role that required school-leaving qualifications in 1980 requires a degree now, and the work is unchanged.
• A profession that adds a master's requirement has raised its entry cost without altering its practice.

Attributed to

Randall Collins

Also called

The credential society; qualification escalation; diploma disease

Standing

Compare social closure, of which this is a mechanism, and effectively maintained inequality.

Page for this entry

Schools reward the cultural habits of advantaged families as though they were ability, so an institution designed to sort by merit sorts partly by origin.

Bourdieu and Passeron's argument is that the school does not need to discriminate to reproduce advantage: it needs only to treat as natural talent the vocabulary, references and manner that some homes supply and others do not. The mechanism is invisible to the institution precisely because it experiences itself as neutral.

In practice

• An admissions process rewarding confident articulacy has measured something households differ in.
• A firm hiring for polish selects on rehearsal opportunity and records it as potential.

Attributed to

Pierre Bourdieu & Jean-Claude Passeron

Also called

Reproduction in education

Standing

Compare concerted cultivation, which describes the parenting practices, and cultural capital, which names the resource.

Further reading

Wikipedia

Page for this entry

Improvement comes from effortful work at the edge of current ability with immediate feedback, not from time served.

Ericsson's studies of violinists found the top group had accumulated far more solitary, effortful, feedback-rich practice — not more playing. He publicly disputed the fixed-hours popularisation that followed.

In practice

• A pianist who plays through pieces for an hour improves slowly; one who works four bars at tempo with a metronome and a recording improves fast.
• A salesperson who makes forty calls learns little; one who reviews three recorded calls with a coach learns a great deal.

Attributed to

K. Anders Ericsson

Also called

The ten-year rule

Possible explanations

The explanation is that improvement requires working at the edge of current ability with immediate feedback, which is effortful and unlike ordinary repetition, so time served predicts little without it.

In its favour

The violin studies found a clear separation by accumulated effortful practice rather than total playing, and structured practice with feedback outperforms unstructured repetition across domains.

Against it

Meta-analysis finds practice explaining a substantial but limited share of performance variance, less in less structured fields such as education and professions, so the strong claim that practice accounts for nearly all of it is not supported.

Standing

The source of the 10,000-hour claim, which Ericsson disputed as a distortion.

Page for this entry

Once quantitative access to a level equalises, advantaged groups preserve their position by securing the qualitatively better tracks, institutions and subjects within it.

Lucas proposed EMI in 2001 as the successor to MMI: when everyone attends, the competition moves inside. Which set, which stream, which subjects, which institution. A system can therefore show complete equality of access and undiminished inequality of outcome, with no rule anywhere having been broken.

In practice

• A school where every pupil takes the same exams sorts them by set, and the sets predict destination better than the exams do.
• A graduate scheme open to all universities recruits overwhelmingly from four.

Attributed to

Samuel Lucas

Also called

EMI

Page for this entry

Believing ability is malleable is claimed to improve persistence and achievement.

Dweck's construct. The replication picture is unflattering: meta-analytic effect sizes on achievement are small, and the large-scale interventions have mostly disappointed.

In practice

• A school replaces "you're clever" with "you worked hard" in every classroom, and the next year's results are indistinguishable from the last.
• A pupil who believed the subject was not for him attempts the harder paper; the pupil already attempting it gains nothing from being told the same thing.

Attributed to

Carol Dweck

Also called

Mindset theory

Possible explanations

The explanation offered is attribution: a pupil who treats ability as fixed reads failure as a verdict and withdraws, while one who treats it as malleable reads failure as information and persists.

In its favour

The attributional mechanism is real and has support in the older literature on learned helplessness, and the construct drew useful attention to how failure is interpreted.

Against it

Meta-analytic effects on achievement are small, several large-scale interventions have disappointed, and the effects that remain are concentrated among lower-attaining pupils rather than general.

Standing

Replication effect sizes are small; the intervention literature is contested.

Page for this entry

Learners produce the expected verbal answer without the model that generates it, and the answer passes because assessment tests the words.

Yudkowsky's example is a pupil who answers that a dropped object falls because of gravity, having attached a word to an event without any expectation about what would happen differently. The diagnostic is anticipation: a person with a model can say what they would expect to see, and a person with a password can only say it again more slowly.

In practice

• A candidate who defines the term correctly and cannot say what would happen if it were false has the password.
• An exam answerable by recombining the lecture's vocabulary will be answered that way by most of the cohort.

Attributed to

Eliezer Yudkowsky

Also called

Password-guessing; answers without models

Standing

Consistent with research on inert knowledge and on transfer failure, both of which document understanding that does not survive a change of surface form.

Page for this entry

Where educational rank is the main route to status, exam competition, private tutoring and household spending on schooling grow beyond the reach of any policy meant to cap them.

Education fever is the Korean name for the intensity with which families pursue schooling, studied by Michael Seth in Education Fever (2002). Its markers are a university entrance examination that sets the national calendar, a private hagwon industry that governments have repeatedly tried and failed to restrain, and Seoul's 10 p.m. curfew on cram schools.

In practice

• A government caps tutoring hours, and the tutoring moves online and into private homes.
• A family moves house for a district's cram schools rather than its state schools.

Attributed to

Korean social concept; studied by Michael Seth

Also called

Education fever

Standing

Scholars link it both to the Confucian examination tradition and to the post-war reality that schooling was the fastest route out of poverty; how much is culture and how much is incentive is debated. Compare shadow education.

Page for this entry

Teaching interventions ranked by averaged effect size, with a threshold below which an intervention is held not to be worth doing.

Visible Learning synthesised hundreds of meta-analyses into a single ranked list with a hinge point around 0.4. Critics including Snook and colleagues and Bergeron have argued that the averaging combines studies whose effect sizes are not comparable, and that some computations do not survive checking. The influence of the list is out of proportion to the confidence its method supports.

In practice

• A school adopts the top three interventions on a list and drops the four beneath a threshold nobody can derive.
• An effect size averaged across reading tests and behaviour scales is a number without a referent.

Attributed to

John Hattie

Also called

Visible Learning; the hinge point

Possible explanations

One explanation for its reach is that it converts hundreds of studies into a single ranked list, which is what a busy school system wants. The hinge point gives an apparently simple rule for what to adopt.

In its favour

It drew attention to practices with good support, including feedback and teacher clarity, and it pushed schools to ask what the evidence says rather than what is fashionable.

Against it

Averaging effect sizes across incommensurable studies and outcomes is heavily criticised, statistical errors have been identified in the underlying syntheses, and the 0.4 threshold has no principled basis.

Standing

Heavily criticised on method: averaging effect sizes across incommensurable studies and outcome measures, and statistical errors identified in the underlying syntheses. The rankings are widely used in policy regardless.

Page for this entry

The return on public investment in people's skills is claimed to be highest in early childhood and to fall steadily with age.

Heckman drew the curve in a 2006 paper in Science from the logic that skills beget skills: capacities built early make later learning cheaper, so money spent at three buys more than money spent at thirty. It became a standard argument for pre-school spending. Rea and Burton tested it in 2020 against a large set of programme cost-benefit estimates and found no sign that earlier programmes paid better.

In practice

• A government moves money from adult retraining to early childhood on the strength of the curve, when its own evaluations show a retraining scheme returning more than any of its pre-school programmes.
• A parent spends freely on toddler enrichment classes on the principle that early spending compounds, then balks at the teenage tutoring that turns out to matter.

Attributed to

James Heckman

Also called

Returns to early investment

Possible explanations

The explanation offered is that skills beget skills, so capacities built early make later learning cheaper and raise the return on everything spent afterwards.

In its favour

The underlying claim about complementarity has support, and several early-childhood programmes with long follow-ups show large returns.

Against it

Rea and Burton found no consistent decline with age across a large set of programme evaluations, and Heckman replied that cost-benefit ratios of dissimilar programmes do not measure the curve, which leaves it well argued and not demonstrated.

Standing

Heckman replied that cost-benefit ratios of dissimilar programmes do not measure the curve, which concerns returns to the same investment at different ages. The theory is widely accepted; the empirical curve has not been demonstrated. Compare the fade-out effect.

Page for this entry

Schools teach, without stating it, a second syllabus of waiting, compliance, deference and being evaluated, and pupils learn it as thoroughly as the first.

Jackson coined the term in Life in Classrooms (1968), after observing that much of a pupil's day is waiting, taking turns, and being praised or judged in public. None of it is on the timetable and all of it is learned. Bowles and Gintis gave it a sharper reading in 1976, arguing that school discipline mirrors the workplace it prepares pupils for.

In practice

• A school that promises independent thinking in its prospectus and rewards silent compliance in its behaviour policy is teaching the second lesson.
• A child learns from years of lining up that whoever controls the queue controls the day.

Attributed to

Philip W. Jackson

Also called

The unwritten curriculum

Standing

Read from different sides: as training in compliance for the labour market, and as the route by which a school's values reach pupils unannounced. Compare guessing the teacher's password.

Page for this entry

Pupils are held to learn better when taught in their preferred modality.

The claim requires an interaction — visual learners doing better with visual instruction and auditory learners worse — and studies designed to detect that interaction do not find it. Surveys nonetheless find most teachers believe it, which makes it one of the most durable beliefs in education without support.

In practice

• A teacher prepares three versions of one lesson and the classes perform identically.
• A pupil told he is a visual learner declines to read the text, which was the part that would have helped.

Attributed to

Various; VARK associated with Neil Fleming

Also called

VARK; visual, auditory, kinaesthetic learners

Possible explanations

One explanation for its persistence is that preferences are real and easily measured, so the leap to matching instruction feels natural. Another is institutional, since the idea offers teachers a way to individualise without changing the curriculum.

In its favour

Pupils do report consistent preferences, and presenting material in several formats is generally useful, which is the practice that survives the theory.

Against it

The claim requires an interaction, with visual learners doing better on visual instruction and worse on other kinds, and studies designed to detect that interaction do not find it.

Standing

The meshing hypothesis has repeatedly failed the test that would support it: matching instruction to stated preference does not improve outcomes. Preferences are real; the benefit of matching them is not.

Further reading

Wikipedia

Page for this entry

Newcomers learn a community's practice by taking part in its real but low-stakes work at the edge, moving towards the centre as their skill and standing grow.

Lave and Wenger's 1991 book Situated Learning drew on studies of apprenticeship among Yucatec midwives, Vai and Gola tailors in Liberia, US Navy quartermasters, supermarket butchers and non-drinking alcoholics in Alcoholics Anonymous. In each, they argued, newcomers learned less by being taught than by taking part in tasks the community really needed, in positions where mistakes were cheap and old-timers were at hand. They presented it as an analytical viewpoint on learning rather than a teaching method.

In practice

• A new editor on a wiki fixes typos and formats citations for months before rewriting an article.
• A junior engineer reviews other people's code and fixes small bugs before owning a feature.

Attributed to

Jean Lave & Etienne Wenger

Also called

LPP; situated learning

Possible explanations

The claim is that learning is a change in a person's participation in a community, so skill and membership grow together. Peripheral tasks are real, which makes the participation legitimate, and low in risk, which makes it learnable. Online it supports the view that lurking can be a form of learning.

In its favour

Preece, Nonnecke and Andrews's 2004 survey of lurkers in online communities found that many said reading was enough for them, and some wanted to get to know the group before taking part.

Against it

It is an analytical lens rather than a testable account of teaching. Fuller and colleagues argued in 2005, from studies of workplace learning, that the model underrates explicit instruction and formal qualifications and that some learners are kept at the periphery rather than being drawn inwards.

Standing

The research behind lurk moar, Rule 33 of the Rules of the Internet. Compare the 90-9-1 Rule and the Zone of Proximal Development.

Further reading

Wikipedia

Page for this entry

Hold the standard fixed and let the time vary, so pupils proceed only once a unit is mastered.

The inversion is of the usual arrangement, where time is fixed by the timetable and attainment varies. Mastery learning fixes attainment and varies time, which requires somewhere for the fast pupils to go and something for the slow ones to be given, and that is where implementations usually fail.

In practice

• A class where nobody advances until everyone can do it produces uniform attainment and a bored third.
• A training scheme that passes people on a date rather than a standard certifies the date.

Attributed to

Benjamin Bloom

Also called

Learning for mastery

Possible explanations

The explanation is prerequisite structure: later material depends on earlier material, so a pupil moved on before mastery accumulates gaps that make everything after it harder.

In its favour

Trials find real gains, particularly for pupils who would otherwise fall behind, and the logic of not building on unlearned foundations is hard to dispute.

Against it

Meta-analytic estimates vary widely with design, and the approach demands flexibility in timetable and staffing that schools rarely have, so measured effects may not transfer to ordinary conditions.

Standing

Meta-analytic estimates of the benefit vary widely with design, and the approach is demanding of timetable and staffing in ways trials often do not reproduce.

Further reading

Wikipedia

Page for this entry

Expanding an educational level reduces class inequality in access to it only once advantaged groups have approached saturation there.

Raftery and Hout proposed MMI in 1993 from Irish data: expansion is absorbed first by the advantaged, so inequality in transition rates persists until their participation nears a ceiling and there is nowhere left for them to expand into. Widening access at a given level therefore does very little until that level is nearly universal for the group already using it.

In practice

• A university expansion that doubles places leaves the participation gap unchanged for a decade, then closes it quickly.
• A training scheme opened to all grades is taken up first by the grades already best served.

Attributed to

Adrian Raftery & Michael Hout

Also called

MMI

Page for this entry

Students admitted to schools where their academic preparation is well below their classmates' may learn less, leave demanding fields, or graduate less often than they would at a school matched to their preparation.

Richard Sander argued it in 2004 in an analysis of American law schools, claiming that admissions preferences left many Black students at schools where they finished lower in the class and passed the bar less often than they would have elsewhere. Arcidiacono, Aucejo and Hotz found in 2016 that minority students at more selective University of California campuses were less likely to finish science degrees than similar students at less selective ones. Bleemer found in 2022 that ending racial preferences in California lowered degree completion and later wages for the students affected, the opposite of what mismatch predicts.

In practice

• A strong student admitted to a highly selective engineering programme finds herself at the bottom of the curve and switches to a major where her grades recover.
• A pupil moved up a set in maths spends the year lost, and does better when moved back to the set that matches her preparation.

Attributed to

Richard Sander

Also called

Academic mismatch

Possible explanations

The proposed explanation is relative preparation: a student placed well below their classmates may struggle, lose confidence and switch away from demanding subjects, where the same student would have persisted elsewhere.

In its favour

Evidence of mismatch is strongest for persistence in science and engineering majors, where studies of the University of California found similar students more likely to finish at less selective campuses.

Against it

Studies of overall graduation and earnings, including the natural experiment created when California ended racial preferences, mostly point the other way, and the hypothesis is politically charged, which has affected how both sides read the data.

Standing

Politically charged and empirically disputed. Evidence of mismatch is strongest for persistence in science and engineering majors; studies of overall graduation and earnings, including the natural experiment in California, mostly point the other way. Compare the big-fish-little-pond effect and the Dale–Krueger finding.

Page for this entry

Intelligence is proposed to comprise several relatively independent modalities rather than one general factor.

Gardner proposed the modalities in 1983 on criteria including neurological isolability and a distinct developmental course. Psychometric work continues to find a substantial general factor that the theory does not predict. Its adoption in schools mostly took the form it least supports, that of teaching each child in their strongest modality.

In practice

• A school labels each pupil with an intelligence and teaches accordingly, which is the application its author disowned.
• A theory whose entry criteria were set by one author is difficult to falsify by adding or removing a modality.

Attributed to

Howard Gardner

Also called

MI theory

Possible explanations

One explanation for its reception is dissatisfaction with reducing ability to a single number. Another is institutional: it arrived when schools were looking for ways to recognise pupils who did poorly on conventional tests.

In its favour

It broadened what schools count as ability, and drew attention to musical, spatial and interpersonal skills that testing had ignored.

Against it

Psychometric work keeps finding a substantial general factor that the theory does not predict, the criteria for admitting a new intelligence are the author's own, and teaching to a pupil's supposed intelligence has not shown benefits.

Standing

The criteria for admitting a modality are Gardner's own and not empirically derived, and the theory has not produced the psychometric independence it predicts. Gardner has himself objected to its use as a theory of learning styles.

Further reading

Wikipedia

Page for this entry

People can do with guidance what they cannot yet do alone, and that gap is where learning happens.

Vygotsky's 1934 concept. Its practical form is scaffolding: support pitched just above current ability and withdrawn as competence arrives, which is why both easy and impossible tasks teach nothing.

In practice

• A trainee given the difficult call with a supervisor listening learns more than one given easy calls alone or hard calls unsupported.
• Both easy and impossible tasks teach nothing, which is why the useful assignment is the one the person cannot quite do yet.

Attributed to

Lev Vygotsky

Also called

Scaffolding; social scaffolding

Further reading

Wikipedia

Page for this entry

A short list of teaching practices drawn from what effective teachers, cognitive science and classroom research agree on.

Rosenshine's 2012 paper is unusual in taking three separate literatures — observational studies of effective teachers, cognitive science on memory, and instructional research — and reporting where they converge. Daily review, small steps, checking understanding, and guided practice appear in all three, which is a stronger warrant than any one of them alone.

In practice

• A lesson that begins with review of last week's material recovers what would otherwise have decayed.
• A teacher who asks four pupils and moves on has sampled, not checked.

Attributed to

Barak Rosenshine

Also called

The ten principles; explicit instruction

Page for this entry

When three walk together, one of them can be my teacher: follow the good in those around you, and correct in yourself the faults you see in them.

Analects VII.22. Confucius goes on to say that he will select his companions' good qualities and follow them, and their bad qualities and avoid them. The saying makes anyone a teacher, by example or by warning, and puts the work of learning on the learner.

In practice

• An apprentice learns as much from watching a careless colleague's mistakes as from the master's method.
• A new manager keeps a note of what her worst boss did, and does the opposite.

Attributed to

Confucius, Analects VII.22

Also called

When three walk together, one can be my teacher; Analects VII.22

Standing

Compare docendo discimus, which runs the same traffic in the other direction.

Page for this entry

A trainee learns a procedure by watching it once, performing it once, and then teaching it to the next trainee.

The maxim is attached to the residency system William Halsted built at Johns Hopkins from the 1890s, which moved surgical training from lectures to graded responsibility on real patients. The phrase itself is later folklore. Its compression of practice into single repetitions has been criticised for decades, and simulation training grew up largely to take the 'do one' off the patient.

In practice

• A hospital lets junior doctors place central lines after watching one, and its complication rate falls only when it adds a mannequin and a checklist before the first real attempt.
• A new barista shown the milk steamer once is training the next starter by the end of the week, and the café's coffee varies accordingly.

Attributed to

Surgical training tradition; associated with William Halsted

Also called

The Halsted model

Possible explanations

The explanation is graded responsibility: the maxim compresses an apprenticeship into three steps, on the assumption that observing, doing and explaining each add something the others cannot.

In its favour

The teaching step is the defensible part, since explaining a procedure improves the explainer's grasp of it, and the residency system it came from was a real improvement on lecture-based training.

Against it

Single repetitions are far short of what skill acquisition requires, patient-safety research documents the cost of learning on real patients, and simulation grew up largely to take the doing step off the patient.

Standing

Criticised on patient-safety grounds and against the evidence that skill needs repeated practice with feedback. The teaching step is the defensible part. Compare deliberate practice and docendo discimus.

Page for this entry

Learning without thought is labour lost; thought without learning is perilous.

Analects II.15, in Legge's translation. The line balances two failures open to a student: accumulating texts without reflecting on them, and reasoning without the knowledge that would check the reasoning. It is among the most quoted Confucian sayings in Chinese education, and the balance it asks for is the one teachers still argue over between knowledge and skills.

In practice

• A course that assesses only recall produces students who know every date and cannot say why any of it mattered.
• A clever student who skips the reading builds an elegant argument on a fact the reading would have corrected.

Attributed to

Confucius, Analects II.15

Also called

Learning without thought is labour lost; Analects II.15

Standing

Compare zhengming, the Analects' other principle about getting the words and the things to match.

Page for this entry

He who can, does; he who cannot, teaches.

Shaw's line comes from the Maxims for Revolutionists appended to Man and Superman (1903), and it became the most quoted insult to teaching in English. Woody Allen's Annie Hall later added a rung further down, for gym teachers. Aristotle had made the opposite case in the Metaphysics: the sign of someone who knows a thing is that he can teach it.

In practice

• A department passes over its best classroom teacher for head of subject in favour of a colleague with more publications.
• A parent tells a child who wants to teach that it is what people do when they cannot do the real thing.

Attributed to

George Bernard Shaw

Also called

He who can, does; Shaw's maxim on teaching

Standing

An epigram rather than a tested claim. Doing and teaching are separate skills that the line treats as one, which is also why the best practitioner is not reliably the best teacher.

Page for this entry

A teacher, or the lesson needed, becomes available once the learner is prepared to receive it.

Usually presented as a Buddhist proverb or credited to Laozi, it has no source in either tradition. The earliest traced form is in Mabel Collins's Theosophical tract Light on the Path in the 1880s, where the disciple and the Master are ready together, and the student-and-teacher version spread through New Thought writing in the early twentieth century. The defensible reading is about attention: readiness changes which of the teachers already present a learner notices.

In practice

• A manager who ignored three offers of mentoring accepts the fourth in the month after a project fails.
• A book that sat unread on a shelf for years becomes the one that makes sense, once the reader has met the problem it answers.

Attributed to

Misattributed to the Buddha and Laozi; Theosophical in origin

Also called

When the student is ready, the teacher will appear; when the pupil is ready, the master appears

Possible explanations

The defensible reading is about attention: readiness changes which of the teachers already present a learner notices, so nothing appears that was not there before.

In its favour

The experience it names is real, since people do report that advice they had heard for years suddenly registers, which fits research on how prior knowledge determines what is taken from instruction.

Against it

It is misattributed, with nothing in Buddhist or Daoist texts corresponding to it, and read literally it encourages waiting for readiness rather than seeking instruction, which reverses the advice most learners need.

Standing

Misattributed. Nothing in Buddhist or Daoist texts corresponds to it; its roots are nineteenth-century Theosophy.

Page for this entry