Education
in a Hurry

The Whole Thing in One Page

The most important event in education is invisible. A pupil listens, copies, answers, practises or submits polished work. None of those acts proves that learning occurred. Learning is the durable change that remains when the teacher, example, notes or software assistance is no longer carrying the difficult part. It may be a fact that can be retrieved, a pattern that can be noticed, a skill that can be performed, or a judgement that can guide a new case.

That change is never produced by delivery alone. Attention must select something from the flood. Working memory must handle it without being overloaded. Existing knowledge must give it somewhere to attach. Practice must make retrieval or action possible after delay. Feedback must expose the error that fluency can hide. Motivation matters because people allocate effort according to expected success, value, cost, trust and belonging. Other people matter because language, imitation, explanation and participation let learners borrow ways of seeing before they can organise them alone.

Teaching arranges those conditions. Sometimes it explains and models. Sometimes it asks a question, narrows a search, demonstrates a movement, supplies a cue or withholds help. The balance changes with knowledge. Guidance that lets a novice begin should be withdrawn as the learner becomes able to plan, choose and correct independently. Difficulty is useful only when it makes the learner perform the mental work that later needs to survive.

Schooling begins with a second problem. Families, communities, workplaces and religious traditions have always educated. A mass society wants millions of children to encounter selected knowledge, during dependable hours, through adults it can appoint and inspect, while also deciding who may proceed to scarce courses, occupations and status. That requires machinery.

Classes let one teacher work with many pupils. Age grades make movement predictable. Timetables divide rooms, adults and hours. Curricula coordinate what strangers in different places are meant to meet. Tests sample what cannot be watched continuously. Grades and credentials turn years of varied experience into portable evidence. No single designer copied this arrangement from a factory. It accumulated through religious instruction, state formation, public demand, urban growth, colonial rule, reform, citizenship and the practical need to coordinate at scale.

The school therefore teaches, supervises, socialises, includes, selects and certifies. These jobs can reinforce one another and can collide. A common examination may weaken patronage while narrowing the curriculum. A standard timetable can make public provision possible while fitting some lives and bodies better than others. A credential may record real capability and still function as a competitive gate.

That is the loop. Because learning is private and delayed, institutions need visible evidence. Once money, progression and identity attach to that evidence, people adapt to it. The proxy starts changing the opportunities, effort and teaching that produce the next round of learning. Good education does not abolish organisation or measurement. It keeps the machinery answerable to the human change for which it exists. The institution earns trust only while its visible success corresponds to capability beyond its walls.

That is the book.

Why You Should Care

In 2024, UNESCO estimated that 273 million children and young people were out of school. That is an access failure on the scale of a large country. It is also an estimate assembled from administrative, survey and population data, and conflict can make the missing harder to count. Yet enrolment cannot settle the question. The last widely cited global learning-poverty update made a different claim: a 2022 modelled simulation suggested that 70 per cent of ten-year-olds in low- and middle-income countries could not read and understand a simple text, compared with a measured pre-pandemic estimate of 57 per cent for 2019. The dates describe different kinds of evidence. The warning survives the distinction. Being in the system and learning from it are separate achievements.

The same distinction follows you beyond school. A page can look familiar until you close it and try to explain the argument. A worked example can make a problem feel obvious until the numbers change. An artificial-intelligence tool can raise the quality of a submission while leaving the user less able to perform without it. These are not failures of character. They are failures to distinguish supported performance from retained capability.

Education research gives a better language for the problem. It explains why prior knowledge can make one person appear quicker, why rereading creates confidence faster than memory, why feedback can guide or smother thought, why a novice and an expert need different forms of help, and why motivation is partly a judgement about expected success, cost and belonging. It also widens the picture. People learn through talk, observation, imitation and participation, not only through private study. A culture supplies the tools, examples and expectations with which an individual mind works.

Then there is the institution that occupied thousands of hours of your life. The classroom feels natural because it arrived before comparison: pupils grouped by age, one adult responsible for the room, time divided by bells, subjects separated on a timetable, progress compressed into marks. None follows directly from a theory of memory. These arrangements solve organisational problems. They coordinate people, buildings, knowledge, care and selection, and they let universities or employers interpret the record of someone they have never met.

The arrangement distributes more than lessons. It distributes adult attention, safety, meals, friendship, discipline, confidence, recommendations, credentials and access to the next gate. Schools can widen opportunity by supplying books, skilled teaching and protected time that a household cannot. They can reproduce advantage when wealth buys tutoring, quiet space, housing near a desired school, cultural familiarity and a second attempt after failure. Both processes can operate in the same building.

This is why educational argument becomes dishonest so quickly. Almost everybody has attended school, so memory arrives disguised as evidence. Reformers take the teacher who transformed them, the lesson that bored them or the child who thrived under one method and scale the case into a system. Politicians prefer changes that can be announced and counted. Technology companies promise individualisation. Traditionalists promise rigour. Progressives promise agency. Each may name a real failure, then mistake one value for the whole institution.

A better set of questions cuts across the camps. What durable change should this arrangement produce? Which mental and social conditions make it likely? What other job is the school being asked to perform? What evidence supports the decision, and what behaviour will the stakes create? Once those questions are separated, some ideological wars become design choices. The remaining conflicts are clearer: who chooses the curriculum, whose knowledge counts, who receives support, and who is selected for what.

Schools are too consequential to romanticise and too useful to dismiss. They are humanity's largest attempt to make access to organised learning a public promise rather than a family accident. Understanding the learner and the machinery together gives you a harder standard for teaching, assessment, reform and your own study.

The Core Ideas

Learning Is a Change, Not a Performance

A lesson is visible. Learning is inferred. The distinction matters whenever somebody must decide whether education worked.

Consider an illustrative classroom problem. A pupil copies a method from the board and completes ten similar questions while the model remains in view. Performance is high. Remove the model a week later, alter the surface details and the method may disappear. The success established that the pupil could follow a supported procedure under those conditions. It did not establish independent command. The same gap appears when a speech can be delivered from prompts but not reconstructed, a piece can be played from notation but not adapted, or a polished answer can be produced with software that carried the reasoning.

Learning means that experience has produced a relatively lasting change in knowledge or capability. The change may concern a fact, concept, movement, perception, strategy, social practice or judgement. Each needs different evidence. Immediate performance still matters, but it comes with a time, setting and level of support attached. Fatigue, anxiety or distraction can depress a performance without erasing previous learning. Recent rehearsal, a helpful cue or an accommodating partner can lift performance without leaving much behind.

Different capabilities leave different traces. A factual claim can be recalled or recognised. A procedure must be executed under changing conditions. Perceptual learning may appear as faster discrimination before the learner can explain the rule. Social learning may appear in participation, judgement and response to others. Evidence should fit the change claimed rather than forcing every outcome through one examination.

Retention is therefore one test, not the whole definition. Can the learner retrieve or perform after a delay? Independence is another. Can the work continue when the prompt, example or expert withdraws? Transfer is the hardest. Can the learner recognise when the knowledge applies in a case that does not advertise the method? No single task settles all three, and the further the new case moves from the original, the more careful the inference must become.

This model blocks two opposite errors. The first is to equate smooth activity with learning. Pupils may be busy, correct and pleased while support conceals dependence. The second is to equate struggle with learning. Confusion can consume an hour and leave the wrong rule more fluent. Productive difficulty makes the learner retrieve, discriminate, explain, decide or correct. Needless difficulty comes from clutter, fear, poor wording, inaccessible materials or knowledge that was assumed but never taught.

Good instruction often improves performance before it tests learning. It explains, models and guides so a novice can enter the task. It then removes assistance and varies the case so the learner must carry more of the process. Good assessment waits long enough and changes enough to discover what remains. That result may look less impressive than a page of perfect answers copied from one pattern. It is a more honest account of capability.

Education begins when activity is judged by the change it leaves behind.

Attention Is the Gate, and Knowledge Changes Its Width

Nothing enters learning merely because it was presented. The learner must notice and process something, and conscious processing has a bottleneck.

Working memory is the limited workspace used to hold and manipulate unfamiliar information now. Its exact capacity depends on the task and the person's knowledge, so one magic number is misleading. The practical constraint is plain. A novice trying to understand a new equation may have to track several symbols, a rule, the purpose of each step and the location of an error at once. Lose one element and the sequence becomes imitation without understanding.

Long-term memory changes the size of the problem. What has been organised into familiar patterns can be treated as a larger unit. A fluent reader does not inspect each letter separately. A skilled driver no longer chooses each hand movement as a new decision. A historian sees an institution where a beginner sees a cluster of unfamiliar names. Expertise appears fast partly because previous learning has compressed the situation. The same design can therefore help a beginner and hinder an expert. Step-by-step prompts reduce search when the route is unfamiliar. Once the route has become organised, the prompts may divide attention and slow performance. Instruction has to change as the learner changes.

This is why knowledge is not the rival of thought. It is part of the machinery that makes thought possible. Critical reasoning about a clinical trial requires concepts such as randomisation, comparison groups and measurement. Searching well requires vocabulary, expectations and enough background to notice when an answer conflicts with the field. Creativity draws on materials already available to be combined, transformed or rejected. A mind can reason with little domain knowledge, but much of its scarce attention must then be spent constructing the problem.

Teaching manages this bottleneck. Clear explanation selects what deserves attention. A worked example displays a route while the learner is still coordinating the pieces. Comparison can reveal which feature changes and which relationship remains. Diagrams help when they make structure easier to see, but can hinder when the learner must split attention across labels, prose and decoration. More information is not always more instruction.

Attention is also social and bodily. Hunger, pain, fear, noise, language demands, sleep loss and the effort of monitoring threat can consume the same limited resource. A pupil who cannot sustain attention may face a badly designed task, an inaccessible environment, a conflict outside the room or a condition requiring specialist support. Treating attention as obedience confuses the visible behaviour with its causes.

The useful question is not whether the learner was exposed to the content. It is what they had to hold in mind, what prior knowledge reduced the burden, and which part of the task received the mental work. A polished slide can direct attention away from the idea. A short, well-chosen example can free enough capacity for the idea to become usable.

Attention opens the gate. Knowledge determines how much can pass through it at once.

Memory Builds the Next Thought

Memory is often pictured as storage and dismissed as the dull part of education. The better model is access. Knowledge matters when it can be reconstructed and used at the moment a new thought needs it.

Familiarity is the easiest substitute. Rereading makes a page quicker to process, and that fluency feels like possession. Closing the page changes the operation. Now the answer must be generated. Retrieval practice uses that gap deliberately. Trying to recall, solve or explain can improve later access more than spending the same time restudying, even when restudy feels easier and produces greater confidence. The attempt practises the act required in the future.

Retrieval need not mean a punishing final examination. It can be a question from last week, a blank page filled before notes reopen, a method reconstructed after the example is covered, or an explanation given without its familiar wording. Feedback matters because an uncorrected error can become easier to repeat. An answer shown before any attempt may be recognised without becoming available for independent recall.

Spacing extends the process through time. Practice packed into one sitting benefits from the answer remaining active. Practice distributed across days or weeks requires the learner to rebuild access after some forgetting. There is no single best interval for every material and retention goal. The stable principle is that important knowledge should return after the previous encounter has cooled, not be exhausted while it is still effortless.

Variation matters when future use requires choosing. Twenty identical percentage questions announce the method through their sequence. A mixed set asks the learner to distinguish a percentage problem from a ratio or rate problem before calculating. Beginners may need a short run of similar examples first. Interleaving is useful when discrimination is part of the target, not as a command to jumble a curriculum.

Prior knowledge gives new material a structure. A reader who knows cricket can connect an innings, a wicket and a declaration; a reader who knows none of them must decode the words and construct the world at the same time. A small, famous study of children's recall for a baseball passage illustrated how topic knowledge can alter comprehension and memory. It does not show that knowledge replaces decoding or general reading skill. It shows that comprehension depends partly on what the reader can bring to the text.

Curriculum matters because knowledge has dependencies. Fractions support ratio, ratio supports rates, and rates support much later work in science and finance. Historical interpretation depends on chronology, institutions and source conventions acquired earlier. A coherent sequence gives new ideas something to connect to and brings important knowledge back before it disappears. It does not require one frozen canon. It requires adults to decide which knowledge gives widest access to further learning, work, culture and public life.

Transfer remains difficult because memory is cued by the conditions in which it was built. A method mastered on a worksheet may not be recognised inside a loan offer. Far transfer across distant domains is rarer than slogans about general problem-solving imply. Transfer improves when teaching names the underlying relationship, compares cases, varies examples and asks the learner to decide when the knowledge applies.

Facts can remain inert when never used. Skills can remain empty when detached from content. Durable learning connects organised knowledge with repeated acts of retrieval, choice and application. Memory is not the opposite of thinking. It is the equipment with which the next thought begins.

People Learn With Other People

A learner's mind changes inside a social world. Before children can study alone, they learn through joint attention, imitation, language, correction and participation in activity that already has a purpose.

A child acquiring a household language does not receive a sequence of grammar lectures. Other people respond to sounds, point, repeat, expand and interpret intentions. An apprentice watches the finished movement, attempts a small part, receives correction and gradually handles more of the task. A musician copies phrasing before being able to explain it. A novice in a laboratory learns where to look and what counts as an anomaly partly by standing beside people who already see the field that way.

This borrowing of attention is one of teaching's deepest powers. An expert can make a feature visible, supply a word for a distinction and demonstrate a sequence that the novice could not yet invent. Conversation also exposes partial understanding. Explaining an idea forces relationships into language. A good question reveals that two people used the same term for different models. Peers can provide examples, challenge and belonging, though group work produces learning only when the intellectual work is distributed rather than quietly handed to the most confident person.

Observation is selective rather than photographic. Learners copy what they attend to, understand and expect to matter. A demonstration therefore needs cues that direct the eye to the decisive feature, and practice that turns watching into doing. Explanation by a peer can help because the peer remembers the novice's difficulty; it can also transmit a confident error. Social learning still needs checking.

The social world supplies motives as well as methods. People notice what their group treats as possible, prestigious or forbidden. A pupil deciding whether effort is worthwhile asks, often without words, whether success is attainable, whether the result has value, what embarrassment may cost and whether people like them belong in the room. Trust makes correction easier to accept. Repeated public failure can make avoidance rational. Warm teacher-pupil relationships are associated with engagement and achievement, but the evidence does not turn affection into a universal treatment. Relationship changes the conditions under which instruction is interpreted.

Learning is also situated in cultural tools and practices. Writing, diagrams, counting systems, musical notation, laboratory instruments and digital interfaces shape the actions a learner can perform. Knowledge acquired in one setting may remain tied to its routines. Participation can produce a highly usable skill without the learner being able to state the rule abstractly; school instruction can produce a stated rule that disappears when the situation looks different. Strong education connects the two.

Social learning has no automatic moral advantage. Communities can preserve error, prejudice and hierarchy with great efficiency. Apprentices may perform labour that benefits a master more than themselves. A group can punish questions and reward conformity. Local participation teaches what the local activity contains, which may be narrow or guarded by caste, class, sex or status.

The conclusion is not that individual memory is an illusion. Nobody can learn on another person's behalf. It is that the individual process is furnished by relationships, language, examples and institutions. The model applies beyond childhood. New employees learn a workplace through tools, routines and correction; athletes absorb timing through demonstration and coached repetition; adults entering an unfamiliar community learn which questions can be asked and which errors are forgiven. The learner eventually carries the capability alone, but rarely builds it alone.

Teaching Transfers the Work

Teaching is the deliberate arrangement of another person's mental and practical work. The argument between teacher-led and child-centred education becomes clearer once the question changes: which decisions should the teacher carry now, and which can the learner carry without losing the learning?

Novices do not yet know which features matter, which steps can be combined or which errors will derail the task. Asking them to discover a complex method with minimal guidance makes working memory search a wide space while trying to understand the result. Reviews of minimally guided instruction have found weak results for beginners in demanding domains. That does not make inquiry or discovery useless. Inquiry works differently when a teacher supplies background knowledge, narrows the problem, sequences evidence, asks diagnostic questions and gives feedback. The learner can make the inference because somebody designed a path on which it becomes reachable.

Direct explanation can also fail. A teacher can speak clearly while pupils process little, or ask questions that one volunteer answers for the room. The useful divide is not explanation against activity. It is whether the learner must attend to the relevant relationship, retrieve what is needed, choose a step, attempt the action and respond to information about error.

Guidance should therefore change. A worked example can show the complete route. The next problem can leave one step blank. Later prompts can become shorter, examples more varied and method choice less obvious. Eventually the learner plans, performs and checks independently. This temporary support is often called scaffolding. The metaphor earns its keep only if the structure comes down. Assistance that remains can become part of the task and conceal dependence. Assistance removed too early turns instruction into a test of prior advantage.

Feedback sits inside the same transfer. A score or the word wrong gives little direction. Useful feedback identifies the gap that matters while leaving the learner able to act. Sometimes the right response is an answer; sometimes a cue, another example, a question or time for a second attempt. Too much feedback performs the diagnosis and correction for the learner. Too little lets the error become fluent.

Motivation changes the amount of work available. Effort is allocated according to expected success, value, cost, belonging and trust. Early success can create interest because competence makes a subject intelligible. Permanent ease can communicate that no growth is expected. Choice can increase ownership when the options are meaningful; unlimited choice can burden a novice who lacks criteria. Growth-mindset messages have small average effects and vary by setting. A slogan about improvement cannot substitute for a credible route to improvement.

Teaching also decides what counts as success. A task that rewards completion encourages different effort from one that rewards revision, explanation or independent use. Public ranking may increase effort for some pupils and make error too costly for others. The teacher cannot control every motive, but can alter the credibility of the goal, the likelihood of progress and the social cost of attempting it.

A classroom adds coordination. Many pupils must listen, practise, wait, speak and risk error in one space. Routines can protect attention by making ordinary actions predictable. Low-stakes checks can reveal pupils hidden by group fluency. Accessibility may require different representations, pacing, language support, assistive technology or ways to respond. Adaptation is strongest when it removes a barrier unrelated to the intellectual goal rather than quietly lowering the goal.

Digital tools belong under the same rule. Video can pause an explanation. Software can offer practice and rapid correction. Generative systems can create hints, examples and finished answers. The crucial question is who performed the operation that needs to remain. A tool may improve the product while removing retrieval, inference or composition. Teaching succeeds when support lets the learner take over the work, not when the support becomes permanently impressive.

Schooling Has Several Jobs

A school is discussed as though it had one purpose and organised as though it had several. Much educational conflict comes from leaving the second fact unstated.

Instruction is the obvious job. Schools create protected time, gather knowledge into curricula and employ adults whose occupation is helping novices acquire it. They can give a child access to books, laboratories, music, languages and mathematical ideas that no household could supply alone. Mass schooling turned some forms of knowledge from inherited privilege into a public expectation.

Supervision and care are also built into the day. Children remain in school while adults work, care and organise the rest of society. Pandemic closures exposed functions that remote content could not replace: meals, safeguarding, routine, specialist support, friendship and a dependable place to be. A reform that changes hours or attendance changes a care system and a labour arrangement, even when its stated aim is instructional.

Schools socialise. They teach explicit content and implicit rules about time, authority, speech, competition, cooperation and whose knowledge receives recognition. Sociologists call the unwritten part the hidden curriculum. Most of it does not require a secret plan. A shared timetable teaches punctuality because rooms and adults are scarce. A uniform may reduce some visible competition while enforcing an authorised presentation of self. Grouping can shape who is expected to lead before anybody writes that expectation down.

Inclusion is another job and another source of conflict. Entering a building does not guarantee access to communication, curriculum, relationships or suitable teaching. Segregated provision can concentrate expertise and remove pupils from the common peer world. Inclusion can widen participation and still fail when the environment creates avoidable sensory, physical, linguistic or instructional barriers. Categories can release support while turning a current need into a fixed identity. The practical question is which barriers the ordinary environment can remove, which individual support preserves participation and when specialist provision is needed.

Schools select and allocate. Places in advanced courses, universities and occupations are scarce or treated as scarce, so institutions need rules for deciding who proceeds. Examinations can replace family connection with a common contest, which is a substantial gain. The contest can still favour pupils whose health, language, previous schooling and private support fit its demands. Merit becomes circular when the institution distributes preparation unequally and then treats the result as pure difference in worth.

Credentials make the judgement portable. An employer cannot watch years of learning, so a certificate compresses performance, persistence and institutional reputation into a signal. Education can raise productivity by building knowledge and skill. Qualifications can also rank applicants or certify traits that selectors cannot observe directly. Both mechanisms can operate in the same degree.

Schools create a peer institution as well. Children compare themselves, form identities, meet lives beyond the household and learn what their group treats as normal. Tracking by prior attainment may narrow the instructional range while changing friendships, expectations and access to the strongest curriculum. A decision presented as pedagogy can become a social pathway.

These jobs collide. Selection demands differentiation; common citizenship demands inclusion. Supervision values dependable schedules; learning may need flexible time. Examination results can focus effort and displace arts, practical work or civic discussion. Families, pupils, universities, employers and governments then argue about quality while measuring different functions.

There is no design that makes the conflict disappear. The honest task is to identify which job a policy serves, which it may damage, who gains and who bears the trade-off. A school can perform several jobs well. It cannot make them the same job.

The Measure Enters the Machine

Learning is private, delayed and uneven. Public systems need to make it visible. Assessment is the bridge, and traffic moves across it in both directions.

A test samples performance under specified conditions. Reliability asks whether the result is sufficiently consistent for the use being made of it. Validity asks whether evidence and theory support the interpretation and decision. A spelling quiz may give useful evidence about the words sampled. It does not measure the whole of literacy. A mathematics examination can assess substantial knowledge and reasoning while remaining sensitive to curriculum coverage, language, time pressure and familiarity with its forms. The number is evidence about a learner, not the learner.

Standardisation makes comparison possible. Common questions and scoring can expose disparities hidden by local confidence, reduce patronage and help a system see whether promised knowledge is being taught. Examinations have sometimes widened access by replacing personal recommendation with performance. Measures become dangerous when the decision exceeds the evidence or the stakes reorganise behaviour around the indicator.

Donald Campbell described the general pressure. The more a quantitative indicator is used for social decision-making, the more people have reason to change their behaviour around it and the more likely it is to distort the process it monitors. In education, a low-stakes quiz can diagnose. Attach school reputation, funding, employment or pupil progression and the tested content attracts time, rehearsal and strategy. Some of that is sensible alignment. Some raises scores without a matching expansion of the wider knowledge the score is taken to represent.

Accountability evidence shows both effects. Test-based systems have improved measured achievement in some subjects and age groups. They have also encouraged curriculum narrowing, threshold gaming and score inflation when gains on the official test do not appear on independent measures. The lesson is not that common testing should disappear. A system without comparable evidence can hide neglect. The lesson is that every measure needs a declared purpose, independent checks and limits on the weight it must carry.

Grades add decisions about task difficulty, timing, resubmission and conduct. Credentials compress the grades again. The Carnegie Unit shows how an imperfect proxy becomes infrastructure. Developed in the early twentieth-century United States, one unit came to represent about 120 hours of study in a subject. Time was countable and portable even though equal time never guaranteed equal learning. Schools, universities, transfer rules and funding arrangements learnt to speak through it, so replacing the measure required a new language among institutions that did not observe one another's teaching.

Metrics also alter identity and opportunity. A rank can tell a pupil where they stand, then change the effort, course, peers and expectations that help determine where they stand next. A diagnostic label can release support and can become a lower curriculum. A prediction becomes an intervention.

The first Core Idea established the problem: learning cannot be read directly from a lesson or a moment of smooth performance. Large systems answer with tests, grades, credits and certificates. Those forms of compression are necessary. Once progression, money and status attach to them, they enter the causal system. They shape teaching, effort, access and the next sample of performance.

The discipline is to reopen the package. What capability does this result support an inference about? Under which conditions and for which decision? What is missing? What behaviour will the stakes invite? Education improves when measurement remains evidence for judgement rather than its replacement.

How It Actually Works

Before the school became ordinary

For most people across most of recorded history, learning was not chiefly a separate institution. Children and novices learnt by joining work, language, ritual, play and communal life.

Across varied societies, recurring methods appear: watching a competent person, imitating part of the action, taking a small responsibility, receiving correction and moving towards work that matters. A learner follows the eye and hand of someone who already knows where to look. Stories preserve classifications and obligations. Games rehearse movement, rules and cooperation. The curriculum is carried in practices rather than printed as a list.

This pattern should not be turned into one romantic human past. Childhood, work and authority differed sharply among foragers, farmers, pastoralists, urban households, castes, guilds and religious communities. Participation can supply purpose, repeated practice and immediate consequences. An apprentice carpenter can see whether a joint fits. A child acquiring a household language receives thousands of responsive exchanges rather than a grammar lecture. The same arrangement can guard knowledge by family, sex, rank or status. A skilled adult may perform expertly and teach badly because the method has become tacit. Apprentices can spend years on labour that benefits the master more than themselves.

Some of the earliest surviving evidence of formal instruction comes from societies that needed specialised knowledge to outlive the person and travel beyond direct imitation. Writing changed the problem. Marks had to be learnt as a system. Records had to be copied accurately. Administrations needed people able to count, classify, correspond and preserve authorised texts.

The first specialist institutions

Mesopotamian tablets preserve school exercises from the early second millennium BCE: sign lists, vocabulary, copied proverbs and calculations, sometimes with the uneven hand of a learner beside a corrected model. The later term edubba, house of tablets, should not be turned into a national school system. Scribal training was narrow, demanding and attached to households, temples and administration. Its importance lies in the new arrangement: a learner practised symbolic work separated from the field or workshop because the skill itself required sustained rehearsal.

Other literate societies built different institutions around different authorities. Egyptian scribes trained for administration and religion. In South Asia, Vedic learning relied on exact oral transmission within restricted traditions, while Buddhist monasteries supported wider scholarly networks. Greek and Roman education combined household training, physical culture, private teachers and schools of rhetoric, with access sharply divided by status and sex.

China produced an unusually long and influential connection between education and public appointment. Confucian texts became central to elite learning, and imperial examinations developed over centuries into a route, never an equal one, towards official status. The mature system demanded mastery of a literary canon and a controlled written form. It could draw talented men beyond hereditary aristocracy, while favouring families able to finance years of preparation and excluding women from the official contest. It demonstrated a possibility that would recur elsewhere: an examination can weaken one inherited privilege while constructing a new educational hierarchy.

Jewish study traditions organised learning around text, interpretation and communal continuity. Christian monasteries, cathedral schools and universities preserved bodies of learning in Europe. Across the Islamic world, homes, mosques, courts, libraries and madrasas supported varied combinations of law, language, philosophy, mathematics, medicine and theology. Authority often lay in teachers and texts rather than one ministry.

Many of these institutions educated minorities. They trained scribes, clergy, officials, merchants and elites, alongside local and religious teaching for wider populations. The examples do not form stages on one road to the modern school. They show different societies separating some kinds of learning from ordinary production when texts, offices, examinations or authorised traditions demanded sustained study. The modern expectation that every child should spend years in a purpose-built institution did not yet exist.

Print, faith and the governing state

Printing lowered the cost of reproducing a stable text, but books do not teach themselves. Early modern religious conflict gave rulers and churches reasons to organise literacy. Protestants wanted believers able to encounter scripture and catechism. Catholic reformers expanded their own schools and teaching orders. States wanted subjects who could receive laws, keep records, serve armies, pay taxes and increasingly share an authorised language.

The relationship was not one-way imposition. Parents and communities often sought schooling because literacy, numeracy and certificates opened work, migration and status. They also resisted fees, distant buildings, unfamiliar languages, harsh discipline or calendars that ignored harvest and household labour. Expansion occurred through bargains among religious bodies, municipalities, employers, philanthropists, families and governments. The balance differed by place.

The state acquired a stronger role because mass provision needs revenue, rules and continuity. Prussia's General School Regulations of 1763 are often treated as the birth certificate of compulsory schooling. They were an important centralising measure, but law on paper exceeded attendance and quality in practice. The same caution applies elsewhere. A statute can establish a duty before buildings, teachers and enforcement exist.

One teacher, hundreds of children

Around 1800, Andrew Bell and Joseph Lancaster promoted versions of the monitorial system in Britain and beyond. One trained adult taught a group of advanced pupils, who then drilled rows of younger learners in reading, writing and arithmetic. In large halls, signals, stations and tightly ordered movements allowed one master to direct hundreds.

The attraction was arithmetic. Cities and reformers wanted basic schooling without paying for one skilled teacher per small class. Monitorial teaching supplied scale through pupil labour. Its weaknesses followed from the same design. The work favoured recitation, copying and routine; monitors knew only a little more than those they taught; surveillance could replace explanation. The system spread rapidly, then lost ground as governments and reformers accepted the cost of trained teachers and age-graded classes.

This episode resembles a factory more than most schooling, yet it did not create the whole modern school. It was one rejected answer to the problem of teaching many people cheaply. The system that replaced it kept the concern with order and scale while investing more authority in the teacher.

The mass school is assembled

From the nineteenth century onward, state-backed mass schooling expanded across much of the world at different speeds and through different bargains. It changed from a minority institution into an expected stage of life for growing populations. No single cause can carry that transformation, and no country supplies the universal sequence.

Industrial and commercial change increased demand for literacy and calculation, but early factories also employed children and could oppose time in school. Urbanisation concentrated pupils and increased demand for supervision. States sought common languages, administrative capacity and citizens who recognised the nation. Religious groups competed for influence. Labour movements and families demanded access to institutions that had become gates to employment and status. Child labour restrictions made attendance easier to enforce, while schooling helped make full-time child labour less legitimate. Cause and consequence fed each other.

The legal landmarks arrived unevenly. In England and Wales, the 1870 Education Act created elected school boards where provision was inadequate; compulsory attendance for children aged five to ten followed in 1880, with fees in elementary schools abolished later. France's Ferry laws of 1881 and 1882 made primary education free and compulsory while strengthening secular state control. Japan's 1872 school order announced a national system during the Meiji transformation, though enrolment, local finance and acceptance took time. Across the United States, the common-school movement expanded tax-supported elementary schooling through state and local politics rather than one federal design.

Many colonial governments exported and imposed forms of schooling for administration, mission, language policy and political control. Colonised communities used the same institutions for mobility, professional formation, anticolonial organisation and claims against the state. Schooling could attack local knowledge and identity while giving pupils the literacy and credentials with which to contest the regime. Its history refuses a clean moral direction.

The organisational form converged because every expanding system faced recurring tasks. It had to place children, allocate teachers, sequence content, record attendance, compare progress and decide who moved on. Age grading gave administrators a predictable flow. A common curriculum made teacher preparation and textbooks easier. The self-contained classroom concentrated responsibility in one adult. Secondary schools divided knowledge among subject specialists. Bells and timetables rationed space and time.

The grammar hardens

Historians David Tyack and William Tobin called these durable arrangements the grammar of schooling. The phrase explains why a classroom divided by age and subject feels like school, while an institution organised around mixed ages, projects or flexible time can appear unserious even when learning is strong. Grammar is not a conspiracy. It is a set of rules that people stop hearing because everyone has learnt to speak through them.

Each rule coordinates expectations beyond the building. Parents plan around the calendar. Teacher contracts, transport, examinations, textbooks, university admissions and public budgets assume certain units of time and progression. Change one piece and the dependencies push back. A school may introduce interdisciplinary projects, then convert the results into subject grades because the next institution requests them. It may promise learning at an individual pace, then move pupils annually because staffing and friendship are age-graded.

The Carnegie Unit shows how a practical standard becomes infrastructure. In the United States, the Carnegie Foundation used a common measure of secondary study when deciding which institutions qualified for its college pension scheme in the early twentieth century. One unit came to mean roughly 120 hours of instruction in a subject. The measure did not claim that every pupil learnt the same amount in those hours. It made preparation comparable enough for admissions, graduation and transfer. Once other decisions used the unit, replacing it required a new common language among institutions that did not trust one another's local judgements.

Reforms often survive by being translated into this grammar. Child-centred activity becomes a scheduled lesson. New technology becomes a screen at every existing desk. Competency-based education becomes outcomes attached to familiar courses. Translation can preserve useful parts, blunt disruptive ones and protect a system from fads. Stability does not prove the form is optimal. It shows that recognisability and coordination have value.

The child enters the theory

The graded school produced a counter-tradition almost as quickly as it produced routines. The reform canon most often taught in Europe and North America begins with Rousseau's Émile in 1762, which imagined education organised around the learner's development rather than the demands of society, though its learner was a fictional privileged boy supervised by a private tutor. Pestalozzi and Froebel turned attention towards activity, objects, affection and early childhood. Froebel's kindergarten treated play as formative work rather than an interval before instruction.

Maria Montessori built a more complete environment. Children chose among carefully sequenced materials, worked with their hands and received immediate information from objects designed to reveal error. Freedom operated inside a prepared order. John Dewey's laboratory school treated education as participation in a community where practical activity could organise inquiry and connect subjects. Neither thinker proposed leaving children alone with whatever interested them.

Developmental psychology then changed the language. Jean Piaget presented children as active constructors of knowledge whose reasoning differed from adult reasoning, though later research found his stage scheme too rigid as a universal timetable. Lev Vygotsky placed language, culture and interaction closer to the centre, emphasising what a learner could do with support from a more capable person. Their work entered schools through compressed slogans, often detached from its limits.

These movements changed early-years provision, classroom materials, teacher education and the legitimacy of pupil activity, first in particular settings and then through international networks, missions, colonial administrations, reform associations and teacher training. Local educators adapted, resisted and combined them with existing practices. They did not dissolve the mass school. Kindergarten became an age grade. Projects received timetable slots. Choice was recorded against curriculum goals. The child entered the theory, while institutions translated the theory into forms they could staff and certify.

Examinations, intelligence and the wider gate

Examinations long predate mass schooling, but mass systems made them a routine technology of movement. Written papers allowed institutions to compare candidates who were not personally known to the examiner. Mark schemes and common syllabuses reduced some patronage and inconsistency. They also encouraged teaching to an examinable canon and made a few performances carry years of consequence.

At the start of the twentieth century, psychological testing promised a more systematic view of difference. Alfred Binet and Théodore Simon developed early scales in France amid arguments about children struggling in compulsory schooling. The textbook story that the ministry hired Binet to invent the test is too neat: he worked through a voluntary society and wider commission, and Simon was indispensable. Their immediate purpose was educational diagnosis. Later users standardised tests, converted results into intelligence quotients and recruited them into military selection and eugenic programmes. A practical instrument became, in some hands, a ranking of persons.

Modern cognitive assessments can predict educational and occupational outcomes to a meaningful degree. They do not reveal a pure substance untouched by language, schooling, health, motivation, familiarity or measurement error, and a score does not set a ceiling. The educational decision remains separate: what follows from the estimate, and what happens if it is wrong?

As secondary schooling expanded, systems had to decide whether pupils should share a common institution or be divided into academic and vocational routes. Early selection can target curricula and protect specialist study. It can also magnify differences in prior opportunity and make later movement difficult. Comprehensive systems delay some decisions but still sort through subject choice, grades, school admissions and informal expectations. Selection does not disappear when a formal track is removed. It changes location.

Higher education expanded from elite formation towards mass participation. Degrees opened professions, developed expertise and offered routes across class boundaries. As participation rose, employers could raise credential requirements without equivalent changes in the work. The achievement became both capability and competitive position. The certificate matters partly because other people are competing for the same gate.

Entitlement, expansion and accountability

After the Second World War, education became more firmly stated as a right and a foundation of development. The 1948 Universal Declaration of Human Rights included free compulsory elementary education and broader access based on merit. Decolonising states expanded schools and universities as instruments of nation-building, public administration, language policy and economic independence. Families treated schooling as a claim on the future. Enrolment rose on an extraordinary scale, though access remained divided by poverty, sex, disability, conflict, location and migration status.

Human-capital theory gave governments an economic explanation: education could raise productivity by building knowledge and skill. Evidence links additional schooling to earnings, though returns vary and no estimate makes every course equally valuable. The language helped justify spending and narrowed attention. Education became easier to defend when it promised growth than when it promised citizenship, culture or autonomy.

By the late twentieth century, many systems had achieved broad attendance and turned towards standards. Governments specified curricula, tested common outcomes, published results and compared schools. England's 1988 reforms created a national curriculum and associated assessment. The OECD's Programme for International Student Assessment began testing samples of fifteen-year-olds in 2000, allowing countries to compare performance in reading, mathematics and science. In the United States, the 2001 No Child Left Behind Act attached escalating consequences to annual test results.

Accountability addressed real failures. Average impressions could conceal groups taught less and expected to fail. Common results exposed disparities and created pressure to improve. They also compressed quality towards what could be tested. Research on No Child Left Behind found mathematics gains for some younger pupils but little evidence of comparable reading gains. A measure can force attention and distort it at the same time.

The screen arrives, and the school remains

Computers repeatedly arrived with a forecast that the classroom would dissolve. Their effects were less theatrical. Schools absorbed computer labs, interactive boards, learning platforms and online homework into existing schedules. The technology changed access and practice where it fitted a clear teaching purpose. It often reproduced a worksheet with better graphics.

Pandemic closures ran the larger experiment nobody wanted. Remote provision protected some continuity, but access to devices, quiet space, adult help and reliable connection was unequal. Many pupils lost learning, and schools' care, safeguarding and social functions became visible through their absence. Content delivery had moved online. The institution had not.

Generative artificial intelligence sharpens the distinction between performance and learning. The OECD's January 2026 review described emerging evidence, not a settled long-run verdict. Systems designed around clear teaching purposes can support learning. General tools can improve completed work without a matching gain once access is removed. Which operation should the system assist, and which must the pupil perform? A machine that supplies every sentence may improve the product while removing the practice.

Schools ended up like this because they solved more than information transfer. New technology can alter explanations, practice, feedback, access and administration. It does not remove the need to decide what should be known, who receives help, how claims are certified and where children spend the day.

How we know

Education research combines experiments, classroom trials, ethnography, longitudinal data, administrative records, international assessments, interviews and archives. Each sees a different object. Laboratory studies can isolate retrieval or cognitive load but may use short tasks unlike a school term. Classroom studies include real teaching and peers but struggle to hold conditions constant. Ethnography can reveal participation and meaning while rarely identifying a clean causal effect. Observational links can inform without proving that one measured feature caused the difference.

Test scores are useful evidence within the knowledge and conditions sampled. They are not interchangeable with learning, wellbeing, citizenship or later income. International comparisons depend on sampling, translation, participation and what each assessment defines as competence. Current global counts often combine administrative records, surveys and population models, so an estimate's observation period and construction matter as much as its publication date. Historical records favour governments, reformers and institutions able to preserve paper; children and families often appear when they resisted or were counted.

The strongest conclusions come from converging methods: learning effects replicated across tasks and settings, institutional patterns supported by history, and policy claims narrowed to the populations studied. There is no single global history of school formation and no laboratory recipe that settles a classroom. Setting-specific evidence remains visibly bounded.

What People Get Wrong

“School was designed as a factory”

The story is irresistible. Industrialists needed obedient workers, copied the factory floor into classrooms, grouped children like batches and installed bells to imitate shifts. It explains every disliked feature at once and provides a villain with a blueprint.

The chronology will not carry it. Formal classrooms, examinations, discipline and standard texts existed long before industrial factories. Nineteenth-century mass schooling did expand beside industrialisation, and both institutions used schedules, hierarchy, specialised roles and records. They faced some of the same coordination problems in growing cities and states. Similar pressures can produce similar forms without one being copied whole from the other.

Schools spread through mixed causes: religious competition, state-building, parental demand, literacy, urbanisation, citizenship, child protection and economic change. Employers sometimes supported education and sometimes preferred cheap child labour. Teachers and reformers often opposed the monitorial halls that most resembled industrial production.

The correction matters because the metaphor directs reform towards appearances. Removing rows, bells or uniforms may change experience, but it does not settle curriculum, teaching, assessment, selection or unequal access. The modern school is an institutional compromise, not a machine whose hidden designer can be exposed and unplugged.

“Everyone has a learning style”

People have preferences. One person likes diagrams, another likes spoken explanation, another prefers to move while working. The unsupported step is the matching claim: that learners classified as visual, auditory or kinaesthetic learn better when the same content is delivered in their preferred mode.

A proper test requires an interaction. Learners must be classified, assigned to different forms of instruction and assessed in a common way. The visual group should outperform with visual teaching while the auditory group outperforms with auditory teaching. Major reviews have found little credible evidence for this crossover pattern.

Content still has a form. Geography often needs maps. Pronunciation needs sound. Dance needs movement. A diagram and explanation together may help many learners because they represent related information, not because each person possesses one channel through which teaching must pass. Disability and language can also require different access routes, which is an argument for accessibility rather than style typing.

The label feels humane because it recognises difference. In practice it can narrow expectation. A pupil told they are not a reader may avoid the practice needed to become one. Ask what representation fits the idea, what barriers this learner faces and whether the result is retained. A personality quiz adds little.

“Knowledge is obsolete when information is searchable”

Search engines and generative systems make retrieval from the outside astonishingly cheap. That changes which details deserve unaided recall. It does not make internal knowledge redundant.

A search begins with vocabulary and judgement. The user must frame the question, distinguish sources, detect missing context and decide whether the answer fits. Without background knowledge, fluent nonsense can pass because nothing in memory resists it. Reading search results also consumes working memory. Knowledge already organised in long-term memory frees attention for comparison and inference.

External tools are excellent for exact dates, tables, rare procedures and updating facts that may have changed. Education should teach when to consult them. It should also build the concepts and reference points needed to use them. A calculator reduces the need for long arithmetic while increasing the value of number sense. Translation software can produce a sentence while leaving the user unable to judge tone or error.

The relevant divide is not memory against technology. It is between knowledge worth carrying because it supports frequent thought and information worth locating when needed. A mind with no stored map cannot tell where the search result belongs.

“The best teaching makes pupils discover everything themselves”

Discovery is memorable because the learner has done intellectual work. The slogan becomes misleading when it treats teacher guidance as contamination.

Beginners do not know the structure they are supposed to find. Open search can overload attention, reward accidental strategies and leave misconceptions intact. Comparisons with guided alternatives often find poor outcomes when beginners must search complex domains with little help. Yet guided inquiry can work well. A teacher can supply background knowledge, narrow the problem, sequence evidence, ask diagnostic questions and withhold the final inference until pupils can make it.

The useful distinction is between mental passivity and instructional guidance. A lecture can require prediction, retrieval and explanation. A project can let one confident pupil think while the rest decorate the poster. Direct explanation can save time that is then spent applying and comparing. Inquiry can reveal why a claim is warranted rather than presenting it as authority.

Expertise changes the balance. Learners should take more control as their knowledge grows. The aim is independence, but independence is an outcome of teaching as well as a method. Asking a novice to behave like an expert does not accelerate the journey. It removes the route map.

“Tests only measure learning”

Tests do measure, imperfectly. They also teach, motivate, sort and change behaviour.

Trying to retrieve an answer can strengthen later access, which is why low-stakes quizzes are useful learning activities. Announced examinations organise study and signal what a system values. Grades can direct attention towards neglected work. These effects are not contamination. They are part of assessment's educational power.

Stakes alter the picture. When progression, employment or school reputation depends on a score, pupils and institutions adapt. More time goes to tested content and formats. Anxiety becomes part of performance. Borderline cases attract strategic attention. Subjects with no metric can lose hours. Scores may rise because knowledge improved, because test technique improved or because the sample became more predictable.

No assessment can cover an entire domain. A result supports an inference of stated width, under stated conditions. Good systems use several forms of evidence, compare trends cautiously and preserve space for outcomes that are hard to standardise. The alternative to testing is not pure understanding. It may be unchecked teacher judgement, patronage or confident ignorance.

A test is both a window and a force applied to the room. Judging it requires looking through it and watching what people do around it.

“Smaller classes always mean better learning”

The intuition is strong: fewer pupils should mean more attention, faster feedback and easier classroom management. Sometimes they do.

Tennessee's Project STAR followed about 11,600 pupils and their teachers, randomly assigning classes from kindergarten to small groups of 13 to 17, regular groups of 22 to 25, or regular groups with an aide. The substantial early-grade reduction produced achievement gains, and later research examined longer-run effects. Random assignment strengthens the causal claim. The result does not prove that every reduction, age, subject or system produces the same return.

Moving from thirty pupils to fifteen changes a teacher's work. Moving from thirty to twenty-eight may change little unless teaching changes with it. Recruitment quality, room, curriculum, behaviour and pupil need matter. A smaller class taught through the same whole-group routine may use its extra capacity weakly. The policy is also expensive because it requires more teachers and rooms, which means less money for pay, training, tutoring, materials or early support.

Class size should therefore be treated as a resource, not a universal quality score. Ask how large the reduction is, where it occurs, which pupils benefit, what teachers will do differently and what the same budget could buy. Arithmetic is part of pedagogy once a whole system has to pay.

“Education is the great equaliser”

Schooling can give a child knowledge, adult attention, meals, friendships and qualifications unavailable at home. It can break inherited limits. This is why excluded groups have fought for access rather than dismissing education as social control.

The same system can reproduce advantage. Families differ in housing, safety, health, language, time, books, networks and capacity to contest a decision. Affluent parents can buy tutoring or move closer to a desired school. Selective institutions can call the resulting performance merit without asking how its prerequisites were distributed. When credentials expand, advantaged groups may acquire more prestigious versions or add further qualifications.

Education therefore affects absolute and relative opportunity differently. Many pupils can become more capable and earn more while the ranking between social groups changes little. Schools cannot repair labour markets, housing, poverty and discrimination alone. Nor are they powerless until every external inequality disappears. Teaching quality, access to demanding curricula, fair discipline, early support and transparent routes can alter lives.

The correction is neither optimism nor fatalism. Education is an equalising institution operating inside an unequal society and performing selection for it. It can widen the field while helping draw the lines. Judge it by who gains real capability, who reaches the next gate and which inherited disadvantages remain priced into the contest.

Use It

Define the change before choosing the activity

Educational plans often begin with an activity. Begin with the result: what should the learner be able to recall, explain, discriminate, make or decide after the support has gone?

The answer changes the design. Recognising a term may need examples and a short check. Choosing a statistical method needs mixed cases so the method is not advertised by the page. Writing an argument needs relevant knowledge, models, attempts, feedback and revision. A lively discussion may help without showing that every participant can later construct the reasoning alone.

State the intended change while accepting that any observation remains a sample. Align explanation, practice and assessment with it. An activity earns its place through what it helps leave behind, not through how modern, enjoyable or demanding it looks.

Find the bottleneck

When a learner stalls, the visible problem is often downstream. A pupil cannot solve the equation because negative numbers remain unstable. A reader cannot interpret the article because half the political institutions are unknown. An employee cannot follow a software procedure because the underlying file structure was never explained.

Work backwards. Is attention being consumed by noise, pain, language, fear or clutter? Is a prerequisite missing? Can the learner recognise the problem but not retrieve the method? Can they perform with a cue but not choose independently? Do they expect any chance of success, and is the result worth the cost?

Motivation often falls after repeated confusion, which makes low effort look like the cause. Diagnose the dependency before judging character. Ask what the successful learner retrieves automatically that the struggling learner must still hold in conscious attention. Repair the smallest unstable part, then return to the larger task quickly enough for the connection to remain visible.

Make recall and use part of learning

Do not wait for a final examination to discover what memory retained. Build retrieval into ordinary work while the cost of failure is low.

Before reopening notes, write what you remember. Answer questions from last week and last term. Cover a worked example and reconstruct the steps. Explain the idea without its familiar wording. Correct the attempt, then return after enough time that recall requires rebuilding rather than recognition.

Practice should resemble the future use. A flashcard can establish a fact; a later problem should require deciding when that fact applies. Mix related methods after the basics are secure enough for discrimination. Change examples, wording and setting so the learner meets the same structure under different cues.

Keep most retrieval low-stakes. Ranking and anxiety can consume attention and discourage honest error. The purpose is information followed by another chance. Memory improves when weakness becomes visible early, not when confidence is preserved until the final surprise.

Transfer responsibility in stages

Judge guidance by what the learner can currently organise, not by loyalty to direct teaching or discovery.

For unfamiliar material, reduce search. Explain the structure, model a process, compare examples and make the decisive feature hard to miss. Ask questions that require the learner to complete part of the reasoning. Then shorten the prompts, vary the problem and require a choice among methods. Later, let the learner plan, create and check with increasing independence.

The sequence avoids two errors. Permanent support produces good-looking performance that collapses when help disappears. Premature independence tests prior advantage and calls the result agency. Scaffolding succeeds when it is adjusted and removed.

Technology follows the same rule. A hint can preserve useful effort. An answer generator can remove it. Ask whether the tool is handling transcription, access or routine feedback, or whether it has taken over the retrieval, inference or composition the learner needs to practise.

Inspect the social conditions

Learning has participants, models and norms. Identify who is doing the thinking, who receives attention, who risks public error and what the group treats as possible.

In group work, give each learner a reason to retrieve, explain or decide before the answers merge. Silent preparation can stop one confident pupil carrying the room. Ask for disagreement and evidence rather than quick consensus. When an expert demonstrates, direct attention to the feature a novice would otherwise miss, then require an attempt.

Treat belonging and trust as conditions, not decorations. A learner who expects ridicule spends attention protecting status. A pupil who never sees someone like them associated with success may treat the subject as another group's territory. Warmth alone does not teach the material, but fair correction, credible expectations and a route to improvement alter whether effort is rational.

Also ask what the setting teaches unintentionally. Timetables, grouping, language rules, discipline and public ranking create a hidden curriculum. They may support order or reproduce an expectation about whose speech counts. The organisation is educating even when no lesson names it.

Match the evidence to the decision

No educational measure deserves to carry every judgement. A test samples some knowledge under stated conditions. A project samples planning and production while admitting different forms of help. Teacher judgement sees patterns across time and can import bias. Attendance reveals access to instruction, not what happened within it. Later outcomes arrive slowly and contain influences beyond school.

Use more than one window when the decision is large. If official scores rise while an independent assessment does not, investigate coaching or narrowing. If coursework is polished while unaided performance weakens, examine how much support entered the product. State what the evidence supports, what it omits and which decision would be too large for it.

Then name the institutional job. Is the policy intended to improve learning, supervise children, widen inclusion, select for a scarce place, reassure parents or certify for employers? One benefit cannot excuse every consequence. Ask who gains, who pays, what behaviour the stakes invite and whether people can recover from an early decision.

The limits

Learning science supplies dependable principles and conditional findings, not one lesson plan. Retrieval helps memory, but the timing, question, feedback and knowledge base matter. Guidance helps novices, while guided inquiry can be excellent and pupils are not novices in everything. Average effects do not settle what one learner needs. Age, subject, language, culture, disability, resources and purpose change the design.

Historical explanation has similar limits. Modern schools share recognisable forms, but there is no single global route to them. A village school, lycée, madrasa and online university may group, teach and certify through different authorities. The model in this book explains recurring pressures without making every system the same.

Education cannot be reduced to measured attainment. Schools shape health, friendship, identity, citizenship, ambition and exposure to art, sport and ways of life. Some outcomes resist clean attribution. That is not permission to praise them without evidence. It is a reason to use several forms of evidence and state value judgements as values.

Learning is not automatically good. People can acquire propaganda, prejudice, manipulation and efficient cruelty. Science can improve educational means. It cannot choose the ends on society's behalf.

The one thing to keep

Keep the change distinct from the machinery.

Learning is the change in a person: knowledge that can be retrieved, a pattern that guides attention, a skill that survives without the model, a judgement that travels to a new case. Schooling is the public machinery built around that change: teachers, peers, classes, calendars, curricula, tests, grades and certificates, carrying several social jobs at once.

The machinery matters. Without public organisation, access to sustained learning is distributed more heavily by family resources, local opportunity and accident. Common institutions can supply skilled help, protected time and entry to knowledge a household cannot provide. Measurement can expose neglect. Credentials can make achievement portable.

The machinery is also persuasive because it is visible. Attendance, hours, marks and degrees can be counted. Human change appears later and is harder to see. That imbalance invites substitution. Time becomes credit. A score becomes ability. A polished submission becomes understanding. A place in school becomes education.

Once you notice the substitution, ask better questions. What did the learner have to think about? What remains when the support is gone? Which relationships and conditions made the work possible? Which school function shaped the decision? What will people do to the measure once opportunity depends on it?

A school earns its claim when its visible machinery keeps producing capabilities that remain useful beyond the machinery. The certificate should open a door. The education should help the person recognise what matters and act after walking through it.

Terms

Learning. Experience producing durable knowledge or capability that remains available later. It is inferred from retention and use, not identified with attendance, activity or immediate performance. Different kinds of learning require different evidence.

Education. The deliberate and accidental formation of knowledge, skill, judgement, identity and values across a lifetime. Schools organise part of education, but families, peers, work and media also educate.

Schooling. Education organised through recognised institutions, roles, timetables, curricula and credentials. The word keeps the social machinery distinct from the learning it is intended to support.

Curriculum. The selected knowledge, practices and experiences learners are expected to encounter, together with their sequence. A curriculum allocates time and therefore makes decisions about importance and exclusion.

Pedagogy. The theory and practice of teaching. It includes how content is represented, how participation is arranged, how understanding is checked and how responsibility moves towards the learner.

Instruction. Deliberate action intended to help someone learn, including explanation, modelling, questioning, practice and feedback. Instruction can be direct or embedded within guided inquiry and discussion. It also designs what the learner must do.

Working memory. The tightly limited system used to hold and manipulate information during conscious thought. Unfamiliar elements compete for this capacity until knowledge makes them easier to organise.

Long-term memory. The system through which knowledge, procedures and experiences remain available beyond the present task. Its organised contents make complex perception, comprehension and reasoning possible.

Schema. An organised knowledge structure that groups related elements and expectations. Schemas let experts recognise patterns quickly, but an inaccurate schema can also guide confident misunderstanding.

Cognitive load. The demand a task places on limited mental processing. Teaching can reduce avoidable load, manage necessary complexity and preserve enough capacity for the learner's useful thinking.

Retrieval practice. Attempting to recall or use knowledge before looking at the answer. Successful retrieval, followed by correction where needed, can strengthen later access more than additional passive exposure.

Spacing. Distributing encounters with material across time. Returning after some forgetting makes memory reconstruct the knowledge and generally supports longer retention than the same practice massed together.

Social learning. Learning through observation, imitation, dialogue, modelling and participation with others. It can direct attention and make practices usable, while also transmitting the errors and hierarchies of the group.

Feedback. Information about the gap between current performance and a goal. Useful feedback directs a change and leaves the learner some thinking to perform rather than replacing the task.

Metacognition. Knowledge and regulation of one's own thinking, including planning, monitoring and checking. Confidence is an imperfect guide, so metacognition benefits from external evidence such as unaided recall.

Transfer. Using learning in a different task, setting or domain. Transfer is commoner when situations share structure and cues, and stronger when varied practice makes the relationship recognisable.

Motivation. The processes that direct, energise and sustain action. In education, effort depends on expected success, value, cost, belonging, relationships and the consequences attached to participation.

Self-regulation. Managing attention, effort, emotion, time and strategy towards a goal. It can be taught and supported, but its demands depend heavily on environment, health and task design.

Scaffolding. Temporary support that allows a learner to perform part of a task not yet manageable alone. Its purpose is to transfer responsibility, so successful scaffolding is adjusted and removed.

Formative assessment. Evidence gathered during learning to decide what should happen next. It becomes formative through use: a quiz, question or observation must lead to adjusted teaching or practice.

Summative assessment. Evidence used to judge achievement at the end of a period or course. It may certify, rank or select, so reliability, validity and consequences become especially important.

Validity. The degree to which evidence and theory support the interpretation and use of an assessment result. A test is not valid in general; a particular inference may be justified.

Reliability. The consistency of a measurement across occasions, raters or comparable items. A result too unstable for an individual decision may still reveal a group pattern when aggregated.

Standardisation. Holding procedures, content or scoring sufficiently constant to support comparison. It can reduce arbitrary variation while narrowing what institutions are willing or able to recognise.

Credential. A recognised certificate of study, competence or status. Credentials make achievement portable between strangers, but can also ration opportunity and acquire value beyond the learning they represent.

Human capital. Knowledge, skill and health treated as productive capacities embodied in people. The concept explains part of education's economic value while omitting civic, cultural and personal purposes.

Signalling. The use of education to communicate traits or relative position to selectors. A qualification can build capability and signal persistence or prior advantage at the same time.

Hidden curriculum. The norms and expectations learnt through school organisation rather than formal content, including relations to time, authority, competition, cooperation, conduct and whose knowledge receives recognition.

Tracking. Dividing pupils into different groups, courses or institutions by estimated attainment, aptitude or destination. Tracking can target teaching while hardening early differences and limiting later movement.

Accountability. Arrangements that make educators or institutions answerable for provision and outcomes. Indicators can expose neglect and focus effort, but high stakes invite adaptation around whatever is counted.

Go Deeper

Daniel T. Willingham, Why Don't Students Like School? (second edition, 2021)

Start here for the learning half. Willingham explains attention, memory, background knowledge, practice and thinking through questions that arise in classrooms rather than through a tour of laboratory terminology. The prose is clear enough for a new reader and precise enough to correct several persistent errors, especially the idea that memory competes with thought. Read it for the connection between what a learner already knows and what that learner can understand next. Its teaching implications are reasoned applications, not rules that erase differences among subjects, ages or pupils.

Barbara Rogoff, The Cultural Nature of Human Development (2003)

Read Rogoff when the individual mind has begun to look too self-contained. She examines learning through participation in families and communities, comparing how children observe, join work, receive guidance and assume responsibility across cultural settings. The book gives social learning more substance than the slogan that people learn together. It is conceptually demanding and its cases cannot represent every community, but it changes the unit of analysis. Instead of asking only what entered one person's memory, it asks how activity, access, expectation and identity organised what became learnable.

Andy Green, Education and State Formation: Europe, East Asia and the USA (second edition, 2013)

For the institutional history, Green shows why national school systems cannot be explained by one factory metaphor or one economic demand. He compares state formation, religion, citizenship, language and development across England, France, Prussia, the United States and East Asia. The coverage remains selective and state-centred, so it should be read beside work on colonial and local educational histories. Its strength is causal discipline: similar school forms emerged through different political bargains, and mass education helped make states and citizens while states were building schools.

David F. Labaree, Someone Has to Fail (2010)

This is the sharpest next book on conflict inside schooling. Labaree argues that American education is pulled among democratic equality, social efficiency and individual mobility. Families may support common provision while seeking positional advantage for their own children, so wider access can coexist with credential competition and hierarchy. The account is deliberately American and should not be universalised carelessly. Read it against Willingham and Rogoff. Better teaching can improve learning, yet institutions may still reward selection, scarcity and signals. That tension explains why educational reform can succeed in one sense and fail in another.

Notes and Sources

Current and changeable material was rechecked on 4 September 2026. Dates attached to international figures refer to the observation or modelled reference period, not merely the year in which a report appeared. The notes identify the evidence supporting the book's main claims and the limits that change how they should be read.

The Whole Thing in One Page and Why You Should Care

The distinction among learning, supported performance and schooling follows the synthesis in How People Learn II and Soderstrom and Bjork's review of learning versus performance. The definition is broad enough to include factual, conceptual, procedural, perceptual and social change. Retention, independence and transfer are treated as different evidential demands rather than interchangeable proof.

UNESCO's current estimate is 273 million children and young people out of school in 2024: 79 million of primary-school age, 64 million of lower-secondary age and 130 million of upper-secondary age. The estimate combines administrative, survey and population data. UNESCO also reports that abrupt conflict-related disruption is poorly captured in several countries and may raise the total by about 13 million, to roughly 286 million. The body uses the published global model estimate and makes the conflict limitation visible.

The learning-poverty numbers require a separate date and measurement warning. The joint 2022 update defined learning poverty as being unable to read and understand a short age-appropriate text by age ten in low- and middle-income countries. Its 57 per cent pre-pandemic estimate was constructed for 2019 from the available assessment and enrolment data. The often repeated 70 per cent figure was a post-pandemic simulation published in 2022, using assumptions about closures and learning loss rather than a new worldwide assessment. The manuscript keeps measured estimate, modelled scenario and publication date separate.

The account of schooling as instruction, supervision, socialisation, inclusion, selection and certification draws especially on Labaree, Green, Tyack and Cuban, Jackson and Meyer, Ramirez and Soysal. These functions vary by place and period. They are an analytic map, not a claim that every institution performs them equally. UNESCO's 2020 inclusion report supports the distinction among placement, participation, support and achievement.

Evidence for the learning model

Learning, capacity and performance. Soderstrom and Bjork review evidence that performance during acquisition can move independently of durable learning. Cowan supports a tightly constrained working-memory model while rejecting one fixed capacity count for every person and task. Long-term knowledge changes the effective unit by organising elements into usable patterns. How People Learn II supplies the wider account of prior knowledge, development, motivation, culture and context.

Retrieval, spacing and transfer. Roediger and Karpicke showed that retrieval can improve delayed retention even when repeated study produces stronger immediate performance. Cepeda and colleagues found a robust spacing advantage across a large experimental literature, with the useful interval depending on how long the learning must last. Dunlosky and colleagues rated practice testing and distributed practice highly but did not offer one universal schedule. Barnett and Ceci's taxonomy supports the manuscript's caution that transfer must specify what changed across task, setting, domain and time.

Prior knowledge. Recht and Leslie's baseball study is a memorable demonstration that topic knowledge can change children's comprehension and recall of a domain-specific passage. It was small and setting-specific. It does not establish that knowledge replaces decoding, language development or general reading instruction. Broader support comes from How People Learn II, Willingham and research on expertise and comprehension.

Social and cultural learning. Rogoff documents guided participation, observation and changing responsibility across varied cultural settings. Vygotsky's work made language, social mediation and supported performance central to later educational thought. These sources do not justify one universal developmental route. The manuscript uses them to correct an individual-only model and keeps power, exclusion and local variation attached to participation.

Motivation, relationships and mindset. Lazowski and Hulleman's meta-analysis found positive average effects from motivation interventions across varied theories and settings, with substantial heterogeneity. Roorda and colleagues found small to medium associations between affective teacher-pupil relationships, engagement and achievement; many underlying studies were observational, so causal claims remain limited. Sisk and colleagues found weak average relationships and intervention effects for growth mindset, while Yeager and colleagues found a small benefit in a large US experiment that varied with pupil and school context. The prose keeps the average modest and the setting visible.

Guidance, inquiry and feedback. Kirschner, Sweller and Clark set out the cognitive case against minimal guidance for novices. Alfieri and colleagues found that unassisted discovery underperformed other approaches while enhanced discovery with feedback or scaffolding could be effective. Furtak and colleagues found positive effects for inquiry-based science teaching, especially where teachers supplied guidance. Wisniewski, Zierer and Hattie found a positive but highly variable average effect for feedback. Together these findings support a changing allocation of responsibility rather than a permanent allegiance to lecture or discovery.

Assessment, incentives and credentials. Campbell's account of indicator corruption concerns the pressure placed on quantitative measures when they carry decisions. The National Research Council reviewed test-based incentive systems and found limited gains alongside strategic responses. Dee and Jacob found mathematics gains for some younger pupils under No Child Left Behind but little evidence of reading gains. Spence supplies the formal signalling model. Angrist and Krueger and Card review or estimate returns associated with additional schooling, while Clark and Martorell found limited evidence for a large diploma signalling effect in one Texas design. Skill formation and signalling are therefore retained as compatible mechanisms whose balance varies by qualification and labour market.

Historical and institutional evidence

Learning before mass schooling. Rogoff supplies the strongest comparative support for learning through observation and participation. The body states recurring mechanisms without treating foragers, farmers, craft households, castes, guilds and religious communities as one social type. Apprenticeship is presented with both its instructional strengths and its capacity to preserve secrecy, hierarchy and exploitation.

Early specialist institutions. Robson supports the account of Mesopotamian scribal exercises and the warning against converting the edubba into a modern national school. Scharfe surveys education in ancient India, while Morgan covers literate education in Greek and Roman settings. Elman establishes the long development, restricted access and social effects of Chinese civil examinations. Berkey supports the description of medieval Islamic education as networks of teachers, texts and institutions rather than one central ministry. These cases are selective comparisons, not stages on one global path.

Print, religion and state formation. Green's comparative history is the main source for the rise of national systems in England, France, Prussia, the United States and East Asia, with different weights given to religion, state formation, language and economic development. Meyer, Ramirez and Soysal document worldwide expansion and convergence from 1870 to 1980. Neither source establishes that one European sequence was copied intact everywhere. The body gives families, local authorities, religious organisations, colonial administrations and colonised communities independent agency.

The English and Welsh landmarks are the Elementary Education Act 1870, which enabled elected school boards to address gaps in provision, and the 1880 Act, which strengthened compulsory attendance for children aged five to ten. Elementary fees were abolished nationally in 1891. France's Ferry legislation made public primary education free in 1881 and compulsory and secular in 1882. Japan's 1872 Gakusei announced a national system, followed by uneven local implementation and repeated revision. In each case, law preceded complete attendance or uniform quality.

Monitorial teaching and the grammar of schooling. Kaestle and Green cover Bell, Lancaster, monitorial instruction and the later move towards trained teachers and graded schools. Tyack and Tobin explain the durability of age grading, self-contained classrooms and time-based credits as a grammar of schooling. Davis, Conroy and Clague examine the limits of the factory metaphor. Similar schedules and hierarchies can arise from shared coordination problems without proving wholesale industrial copying.

Silva, White and Toch trace the Carnegie Unit, introduced in the United States in the early twentieth century and commonly expressed as about 120 hours of instruction in a secondary-school subject. The manuscript treats it as a time-based currency for coordination, preparation and transfer, not a direct measure of equal learning.

Child-centred and developmental traditions. Rousseau, Pestalozzi and Froebel are treated through the later histories cited by Green and Tyack and Cuban. Dewey and Montessori supply primary statements of purposeful activity, community, prepared environments and guided independence. Piaget's work established a developmental constructivist programme whose fixed stage interpretations have been revised substantially. Vygotsky's posthumously assembled English-language collection shaped later accounts of social mediation and supported performance. Their ideas moved through local adaptations rather than intact implementation.

Testing and selection. Binet and Simon developed their scales in the context of identifying children who might need different educational help. Brysbaert and Nicolas correct two textbook myths: the French government did not commission Binet alone to invent the test, and Théodore Simon's collaboration was central. Carson traces how intelligence measurement became tied to merit, selection and hereditarian claims in France and the United States. Deary reviews predictive evidence while retaining measurement error and environmental influence. The manuscript separates prediction from fixed worth and from the institutional decision made after testing.

Schofer and Meyer document the expansion of higher education. Article 26 of the 1948 Universal Declaration of Human Rights supplies the right-to-education language. Green and Seth support the account of schooling under state formation and colonial rule, including administrative uses and appropriation by colonised groups. Outcomes varied by empire, language, region, local politics and access.

Accountability and current technology. England's Education Reform Act 1988 established the national curriculum and associated assessment framework. PISA began in 2000 and assesses samples of fifteen-year-olds in reading, mathematics and science; it does not test every pupil or represent an entire national curriculum. The No Child Left Behind account uses the Act's annual-testing structure and the evaluations above.

The current artificial-intelligence discussion relies on the OECD's Digital Education Outlook 2026, published on 19 January 2026. The OECD describes emerging evidence that purpose-built or pedagogically guided systems can support learning, while general-purpose systems may improve completed work without a matching gain when access is removed. The manuscript states a design question and no settled long-run effect size.

What People Get Wrong and Use It

The learning-styles correction follows Pashler and colleagues, who asked whether matching instruction to an assessed preference produced the required crossover interaction and found inadequate evidence. The text keeps that claim separate from accessibility, preference and choosing a representation suited to the content.

The class-size account is anchored in Tennessee's Project STAR and Krueger's experimental analysis. About 11,600 pupils entered the study in kindergarten and were randomly assigned to small classes of roughly 13 to 17 pupils, regular classes of roughly 22 to 25, or regular classes with an aide. This design strengthens causal inference within early primary grades in that programme. It does not establish equal effects for minor reductions, older pupils or systems unable to recruit additional teachers of comparable quality.

Chetty, Friedman and Rockoff connect teacher value-added estimates in one large US district to later outcomes. Jackson shows that teacher effects on test scores and non-test school outcomes are weakly correlated and that both can predict later consequences. The findings support the claims that teachers matter and that one score captures only part of their work. They do not justify ranking every teacher through one model.

The equality discussion is intentionally two-sided. Expanded schooling can build capability, increase earnings and widen access while competitive credentials preserve relative hierarchy. Labaree supplies the clearest institutional statement of that tension. The practical lenses are deductions from the evidence already identified, not claims of one best method for every learner or system.

Bibliography

Primary and official sources

Binet, Alfred, and Théodore Simon. The Development of Intelligence in Children: The Binet-Simon Scale. Translated by Elizabeth S. Kite. Baltimore: Williams & Wilkins, 1916.

Dewey, John. Democracy and Education: An Introduction to the Philosophy of Education. New York: Macmillan, 1916.

Montessori, Maria. The Montessori Method. Translated by Anne E. George. New York: Frederick A. Stokes, 1912.

Piaget, Jean. The Origins of Intelligence in Children. Translated by Margaret Cook. New York: International Universities Press, 1952.

Vygotsky, L. S. Mind in Society: The Development of Higher Psychological Processes. Edited by Michael Cole, Vera John-Steiner, Sylvia Scribner, and Ellen Souberman. Cambridge, MA: Harvard University Press, 1978.

National Academies of Sciences, Engineering, and Medicine. How People Learn II: Learners, Contexts, and Cultures. Washington, DC: National Academies Press, 2018.

National Research Council. Incentives and Test-Based Accountability in Education. Washington, DC: National Academies Press, 2011.

OECD. PISA 2022 Results, Volume I: The State of Learning and Equity in Education. Paris: OECD Publishing, 2023.

OECD. OECD Digital Education Outlook 2026: Exploring Effective Uses of Generative AI in Education. Paris: OECD Publishing, 2026.

UNESCO. Global Education Monitoring Report 2020: Inclusion and Education, All Means All. Paris: UNESCO, 2020.

Global Education Monitoring Report Team. 2026 Global Education Monitoring Report: Access and Equity. Countdown to 2030. Paris: UNESCO, 2026.

United Kingdom. Elementary Education Act 1870. London, 1870.

United Kingdom. Elementary Education Act 1880. London, 1880.

United Kingdom. Education Reform Act 1988. London: HMSO, 1988.

United Nations. Universal Declaration of Human Rights. Paris, 1948.

World Bank, UNESCO, UNICEF, Foreign, Commonwealth and Development Office, USAID, and Bill & Melinda Gates Foundation. The State of Global Learning Poverty: 2022 Update. Washington, DC: World Bank, 2022.

Modern works and research

Alfieri, Louis, Patricia J. Brooks, Naomi J. Aldrich, and Harriet R. Tenenbaum. “Does Discovery-Based Instruction Enhance Learning?” Journal of Educational Psychology 103, no. 1 (2011): 1-18.

Angrist, Joshua D., and Alan B. Krueger. “Does Compulsory School Attendance Affect Schooling and Earnings?” Quarterly Journal of Economics 106, no. 4 (1991): 979-1014.

Barnett, Susan M., and Stephen J. Ceci. “When and Where Do We Apply What We Learn? A Taxonomy for Far Transfer.” Psychological Bulletin 128, no. 4 (2002): 612-637.

Berkey, Jonathan P. The Transmission of Knowledge in Medieval Cairo: A Social History of Islamic Education. Princeton: Princeton University Press, 1992.

Brysbaert, Marc, and Serge Nicolas. “Two Persistent Myths About Binet and the Beginnings of Intelligence Tests in Psychology Textbooks.” Collabra: Psychology 10, no. 1 (2024): 117600.

Campbell, Donald T. “Assessing the Impact of Planned Social Change.” Evaluation and Program Planning 2, no. 1 (1979): 67-90.

Card, David. “The Causal Effect of Education on Earnings.” In Handbook of Labor Economics, volume 3A, edited by Orley Ashenfelter and David Card, 1801-1863. Amsterdam: Elsevier, 1999.

Carson, John. The Measure of Merit: Talents, Intelligence, and Inequality in the French and American Republics, 1750-1940. Princeton: Princeton University Press, 2007.

Cepeda, Nicholas J., Harold Pashler, Edward Vul, John T. Wixted, and Doug Rohrer. “Distributed Practice in Verbal Recall Tasks: A Review and Quantitative Synthesis.” Psychological Bulletin 132, no. 3 (2006): 354-380.

Chetty, Raj, John N. Friedman, and Jonah E. Rockoff. “Measuring the Impacts of Teachers II: Teacher Value-Added and Student Outcomes in Adulthood.” American Economic Review 104, no. 9 (2014): 2633-2679.

Clark, Damon, and Paco Martorell. “The Signaling Value of a High School Diploma.” Journal of Political Economy 122, no. 2 (2014): 282-318.

Cowan, Nelson. “The Magical Number 4 in Short-Term Memory: A Reconsideration of Mental Storage Capacity.” Behavioral and Brain Sciences 24, no. 1 (2001): 87-114.

Davis, Robert A., James C. Conroy, and Julie Clague. “Schools as Factories: The Limits of a Metaphor.” Journal of Philosophy of Education 54, no. 5 (2020): 1471-1488.

Deary, Ian J. “Intelligence.” Annual Review of Psychology 63 (2012): 453-482.

Dee, Thomas S., and Brian Jacob. “The Impact of No Child Left Behind on Student Achievement.” Journal of Policy Analysis and Management 30, no. 3 (2011): 418-446.

Dunlosky, John, Katherine A. Rawson, Elizabeth J. Marsh, Mitchell J. Nathan, and Daniel T. Willingham. “Improving Students' Learning with Effective Learning Techniques: Promising Directions from Cognitive and Educational Psychology.” Psychological Science in the Public Interest 14, no. 1 (2013): 4-58.

Elman, Benjamin A. A Cultural History of Civil Examinations in Late Imperial China. Berkeley: University of California Press, 2000.

Furtak, Erin Marie, Tina Seidel, Heidi Iverson, and Derek C. Briggs. “Experimental and Quasi-Experimental Studies of Inquiry-Based Science Teaching: A Meta-Analysis.” Review of Educational Research 82, no. 3 (2012): 300-329.

Green, Andy. Education and State Formation: Europe, East Asia and the USA. Revised and expanded second edition. Basingstoke: Palgrave Macmillan, 2013.

Jackson, C. Kirabo. “What Do Test Scores Miss? The Importance of Teacher Effects on Non-Test Score Outcomes.” Journal of Political Economy 126, no. 5 (2018): 2072-2107.

Jackson, Philip W. Life in Classrooms. New York: Holt, Rinehart and Winston, 1968.

Kaestle, Carl F. Pillars of the Republic: Common Schools and American Society, 1780-1860. New York: Hill and Wang, 1983.

Kirschner, Paul A., John Sweller, and Richard E. Clark. “Why Minimal Guidance During Instruction Does Not Work: An Analysis of the Failure of Constructivist, Discovery, Problem-Based, Experiential, and Inquiry-Based Teaching.” Educational Psychologist 41, no. 2 (2006): 75-86.

Krueger, Alan B. “Experimental Estimates of Education Production Functions.” Quarterly Journal of Economics 114, no. 2 (1999): 497-532.

Labaree, David F. Someone Has to Fail: The Zero-Sum Game of Public Schooling. Cambridge, MA: Harvard University Press, 2010.

Lazowski, Rory A., and Chris S. Hulleman. “Motivation Interventions in Education: A Meta-Analytic Review.” Review of Educational Research 86, no. 2 (2016): 602-640.

Meyer, John W., Francisco O. Ramirez, and Yasemin Nuhoglu Soysal. “World Expansion of Mass Education, 1870-1980.” Sociology of Education 65, no. 2 (1992): 128-149.

Morgan, Teresa. Literate Education in the Hellenistic and Roman Worlds. Cambridge: Cambridge University Press, 1998.

Pashler, Harold, Mark McDaniel, Doug Rohrer, and Robert Bjork. “Learning Styles: Concepts and Evidence.” Psychological Science in the Public Interest 9, no. 3 (2008): 105-119.

Recht, Donna R., and Lauren Leslie. “Effect of Prior Knowledge on Good and Poor Readers' Memory of Text.” Journal of Educational Psychology 80, no. 1 (1988): 16-20.

Robson, Eleanor. Mathematics in Ancient Iraq: A Social History. Princeton: Princeton University Press, 2008.

Roediger, Henry L. III, and Jeffrey D. Karpicke. “Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention.” Psychological Science 17, no. 3 (2006): 249-255.

Rogoff, Barbara. The Cultural Nature of Human Development. New York: Oxford University Press, 2003.

Roorda, Debora L., Helma M. Y. Koomen, Jantine L. Spilt, and Frans J. Oort. “The Influence of Affective Teacher-Student Relationships on Students' School Engagement and Achievement: A Meta-Analytic Approach.” Review of Educational Research 81, no. 4 (2011): 493-529.

Scharfe, Hartmut. Education in Ancient India. Leiden: Brill, 2002.

Schofer, Evan, and John W. Meyer. “The Worldwide Expansion of Higher Education in the Twentieth Century.” American Sociological Review 70, no. 6 (2005): 898-920.

Seth, Sanjay. Subject Lessons: The Western Education of Colonial India. Durham, NC: Duke University Press, 2007.

Silva, Elena, Taylor White, and Thomas Toch. The Carnegie Unit: A Century-Old Standard in a Changing Education Landscape. Stanford, CA: Carnegie Foundation for the Advancement of Teaching, 2015.

Sisk, Victoria F., Alexander P. Burgoyne, Jingze Sun, Jennifer L. Butler, and Brooke N. Macnamara. “To What Extent and Under Which Circumstances Are Growth Mind-Sets Important to Academic Achievement? Two Meta-Analyses.” Psychological Science 29, no. 4 (2018): 549-571.

Soderstrom, Nicholas C., and Robert A. Bjork. “Learning Versus Performance: An Integrative Review.” Perspectives on Psychological Science 10, no. 2 (2015): 176-199.

Spence, Michael. “Job Market Signaling.” Quarterly Journal of Economics 87, no. 3 (1973): 355-374.

Tyack, David, and Larry Cuban. Tinkering Toward Utopia: A Century of Public School Reform. Cambridge, MA: Harvard University Press, 1995.

Tyack, David, and William Tobin. “The Grammar of Schooling: Why Has It Been So Hard to Change?” American Educational Research Journal 31, no. 3 (1994): 453-479.

Willingham, Daniel T. Why Don't Students Like School? A Cognitive Scientist Answers Questions About How the Mind Works and What It Means for the Classroom. Second edition. San Francisco: Jossey-Bass, 2021.

Wisniewski, Benedikt, Klaus Zierer, and John Hattie. “The Power of Feedback Revisited: A Meta-Analysis of Educational Feedback Research.” Frontiers in Psychology 10 (2020): 3087.

Yeager, David S., et al. “A National Experiment Reveals Where a Growth Mindset Improves Achievement.” Nature 573 (2019): 364-369.