The Gold Standard of History of Educational Psychology

Chapter index

Teaching existed long before educational psychology became a named research field. People could demonstrate a craft, explain a text or correct a calculation without possessing an experimental account of memory. The historical change examined here is more specific: teaching claims increasingly became questions about learners, mechanisms, comparisons and evidence that could be disputed publicly.

Consider the claim that copying a rule ten times makes it understood. Four different propositions are hidden inside it. The learner may reproduce the words accurately, remember them tomorrow, explain their meaning or select the rule in an unfamiliar problem. Evidence for the first proposition does not automatically establish the other three. Educational psychology becomes useful when it separates these outcomes rather than allowing one convenient measurement to represent them all.

History also asks why a particular outcome mattered. A classroom organised around examination rankings may privilege speed and accuracy. A teacher concerned with scientific inquiry may also value question formation and revision of explanations. Neither the instrument nor its score tells the whole story of the educational purpose. We must inspect the problem an investigator believed needed solving.

A selective chronology, not a single succession

The landmarks used here are James’s 1899 talks, Thorndike and Woodworth’s 1901 transfer investigations, Dewey’s 1910 account of reflective thought, Skinner’s 1958 teaching-machine paper, Piaget’s developmental discussions published in 1964, Sweller’s 1988 cognitive-load paper, and evidence syntheses from 2007, 2025 and 2026. These are dated publications, not dates on which their subjects suddenly began.

This is a selective route through predominantly European and American sources, not a complete global history. It must not imply that educational knowledge originated only in these settings, or that each later author displaced every earlier approach. A theory, a teaching practice and an institution can develop on different timetables. Comparing their questions is more informative than arranging famous names as a parade of inevitable progress.

A useful reading habit is to complete this sentence for every source: the author studied or argued for this outcome, using this evidence, under these conditions. If one of those parts is missing, the historical account is not yet ready to support a classroom conclusion.

Contents · Next chapter

In Talks to Teachers on Psychology, first published in 1899, James addressed the relationship between psychological knowledge and teaching judgment. His preface traces the talks to lectures requested in 1892. The digitised copy linked here carries a later imprint; the publication history should not be confused with the date of its digital release.

James resisted the expectation that a science of mind could directly generate one correct classroom programme. Psychological principles could constrain and illuminate teaching, but moving from general principles to a particular lesson required practical judgment. Several methods might be compatible with the same psychology. This is an argument about the relationship between disciplines, not a controlled experiment comparing teaching methods.

The distinction remains useful when reading historical educational advice. An explanation of attention does not, by itself, prescribe a lesson’s topic, examples, timing or assessment. Those decisions involve the learner’s existing knowledge and the purpose of the activity. The scientific claim and the instructional choice need a connecting argument.

Worked interpretation: one principle, two lessons

Suppose two teachers want pupils to understand equivalent fractions. One begins with a divided rectangle; the other begins with numerical examples and then asks pupils to draw them. Both may aim to connect unfamiliar relationships with familiar knowledge. Their shared aim does not prove that both designs will work equally well, or that either will work for every class.

To compare them, specify the outcome before declaring a winner. If the goal is recognition, ask pupils to identify equivalent pairs. If the goal is explanation, ask why multiplying numerator and denominator by the same nonzero number preserves the fraction’s value. If the goal is transfer, change the representation and remove the original cue. The lesson’s compatibility with a principle is an initial justification; learner evidence must complete the evaluation.

James’s caution therefore helps prevent two opposite mistakes: treating psychological research as irrelevant to teaching, and treating a general psychological statement as a finished lesson plan. A historically careful reader can preserve the distinction without endorsing every assumption elsewhere in his book.

Previous chapter · Contents · Next chapter

James’s discussions of interest and attention show why visible participation can mislead an observer. His treatment connects acquired interest with existing concerns and associations, and distinguishes attracting attention from sustaining useful mental work. Some of his generalisations about childhood reflect his period and should be read critically rather than adopted as universal developmental facts.

The historical issue is precise: when a learner looks interested, what object has become interesting? A demonstration can attract attention to a surprising event while leaving its explanation poorly understood. A pupil may remember the colour, joke or dramatic moment but not the relation that the teacher intended to teach. Engagement is evidence that something captured attention; identifying what was learned requires another observation.

An original demonstration case

Imagine a lesson using two toy vehicles to introduce speed. The vehicles are visually different. Pupils are asked which is faster, and most select the brighter vehicle after an exciting race. The teacher has not yet established whether pupils compared distance and time or simply remembered which vehicle looked more impressive.

Replace the spectacle with a small calculation. Vehicle A travels 6 metres in 3 seconds; vehicle B travels 8 metres in 4 seconds. Under the simplifying assumption that average speed is the target, both average 2 metres per second. Greater distance alone does not make B faster. Ask pupils to explain the equality before showing any new race.

Then change the colours and labels. If the judgment changes while the distances and times do not, the irrelevant cue is influencing the answer. This is a diagnostic teaching example, not an experiment establishing a general law of attention. It gives the teacher a more discriminating question than whether the class enjoyed the activity.

The repair is to direct attention to the intended relation, ask learners to use it, and check it with another representation. The historical contribution is not a recipe for entertainment. It is the invitation to investigate the content of attention and the connection between an attractive experience and an educational outcome.

Previous chapter · Contents · Next chapter

Thorndike and Woodworth’s 1901 paper on improvement in one mental function and performance in others examined whether training carried beyond the particular activity practised. Their examples included estimating magnitudes and identifying specified features in printed material. They challenged the idea that broad labels such as attention or accuracy necessarily named one general capacity improved uniformly by exercise.

Their method matters as much as their conclusion: measure performance on tasks, train a selected activity, and investigate performance on other tasks. They also distinguished improved performance from an improved capacity to learn a new task. Those are different possible consequences of training. Their account associated transfer with shared elements, but the historical proposal should not be converted into a complete modern theory of all transfer.

Worked calculation: gains are outcome-specific

Consider invented scores for a learner who practises estimating rectangle areas. On a 20-item rectangle test, the score rises from 8 to 16. On a separate 20-item irregular-shape test, it rises from 7 to 9. The rectangle gain is 8 points, or 40 percentage points of the available score. The irregular-shape gain is 2 points, or 10 percentage points.

The two gains show that performance changed differently on these measures. They do not yet show that training caused either gain. Repeating a test, receiving outside instruction or becoming more familiar with the response format could also contribute. A comparison condition and a careful account of what happened between tests would strengthen the inference.

Nor does the smaller second gain prove that transfer is impossible. Perhaps the new shapes require decomposition that was never taught. Perhaps the test is too short to measure a small change reliably. Perhaps the learner can explain a transferable principle but has not yet used it fluently. Each possibility suggests a different follow-up.

This landmark makes a durable demand on educational claims: name the destination of learning. Improving one performance is valuable, but it is not permission to claim improvement across every activity sharing a broad psychological label. The Education parent develops the complementary aim of learning that transfers into life.

Previous chapter · Contents · Next chapter

In How We Think, published in 1910, Dewey distinguished reflective inquiry from the mere succession of ideas or the acceptance of a belief without examining its grounds. His concern was the quality of thought: how a difficulty becomes a question, how possible explanations arise, and how their implications are checked.

This is not adequately represented by the slogan that pupils should always discover everything alone. A problem can be poorly framed, relevant information can be unavailable, and activity can proceed without inquiry. The educational question is whether the learner is examining reasons and using evidence to settle a difficulty. A room full of movement does not establish that this has happened.

Worked inquiry: two explanations for a reading error

A learner reads the sentence “The bat flew out of the cave” and explains it as a piece of sports equipment moving through the air. Two hypotheses are available. The learner may know only the sporting meaning of bat, or may know the animal meaning but fail to use context to select it.

A useful next question distinguishes those hypotheses: ask the learner to describe an animal called a bat without referring to the sentence. If that meaning is unavailable, explain and illustrate it before asking for contextual selection. If it is available, compare sentences using the two meanings and ask which surrounding words support each interpretation. The same initial error can therefore lead to different teaching decisions.

The answer should remain open to revision. A learner who gives the expected definition may still struggle with the sentence for another reason, such as unfamiliar vocabulary elsewhere. The teacher’s hypothesis is a working explanation, not an identity assigned to the learner.

This case adapts the logic of inquiry; it is not a diagnostic instrument or a claim about Dewey’s own classroom data. It shows why reflective teaching requires more than a correction. The teacher asks what would make competing explanations distinguishable, obtains evidence, and changes the response accordingly. Historical theory becomes practically useful when it improves that reasoning rather than supplying a fashionable label.

Previous chapter · Contents · Next chapter

Dewey’s discussion of meaning and language challenges a familiar classroom shortcut: accepting a polished verbal formula as proof that an idea has become clear. A definition can be accurately repeated while the learner cannot recognise a case, explain a boundary or apply the relation. The historical question concerns what words are doing in thought, not whether vocabulary should be taught.

To investigate understanding, vary the demand. A learner may define average speed as distance divided by time. Ask for a calculation, then ask whether averaging two speeds always gives the average speed of a journey. The second question exposes an assumption hidden by the fluent definition.

Worked example: the average of speeds is not always journey speed

Assume a traveller covers 60 kilometres at 30 kilometres per hour and another 60 kilometres at 60 kilometres per hour, with no stops. The first leg takes 60 ÷ 30 = 2 hours. The second takes 60 ÷ 60 = 1 hour. Total distance is 120 kilometres and total time is 3 hours, so the journey’s average speed is 40 kilometres per hour.

The arithmetic mean of the two speeds is 45 kilometres per hour. It is not the answer because the traveller spends unequal amounts of time at the two speeds. If instead each speed were maintained for one hour, total distance would be 30 + 60 = 90 kilometres over 2 hours, giving 45 kilometres per hour. The changed condition changes which calculation is appropriate.

A learner who initially answers 45 need not be labelled careless. Ask what quantity is being averaged, what the equal weights represent and whether the journey provides those weights. The repair is a comparison between the two cases, followed by a fresh problem in which the learner identifies the relevant totals independently.

This is one way to turn a historical concern about verbalism into a concrete teaching check. The definition remains useful, but it must organise the calculation and the explanation. Without that connection, fluent language can conceal a weak distinction precisely where the learner most needs it.

Previous chapter · Contents · Next chapter

Skinner’s Teaching Machines, published in Science in 1958, distinguished equipment that presents material from arrangements that require learner responses. He discussed Sidney Pressey’s earlier devices, immediate information about answers and individual pacing. These concerns predate today’s digital learning platforms.

An especially revealing qualification appears later in the paper: Skinner acknowledged the need to prepare learners to work without the machine’s helpful conditions. He also stated that a large-scale evaluation of machine teaching had not yet been attempted. The article is therefore evidence of a programme, design reasoning and early experience, not proof that all its anticipated benefits had been established.

Original audit of a practice device

Suppose a device gives a student 40 arithmetic prompts and displays whether each response is correct. The student answers 36 correctly while using the device, then answers 12 of 20 new problems correctly without it the next day. The first accuracy is 90%; the second is 60%. The denominator and the task have both changed, so report both rather than subtracting raw counts.

The difference does not by itself identify the cause. The new problems may be harder; the delay may matter; the device may have provided cues; or the first score may include repeated attempts. Start by documenting what “correct” meant. Did it mean first response, eventual response or response after a hint? These are materially different records.

A better audit would preserve first-attempt results, record assistance, include comparable independent items and add a delayed assessment. If the aim is flexible calculation, also ask students to explain a method and choose among operations. A high rate of completion is useful operational information, but it cannot substitute for evidence about the intended learning.

The historical comparison is especially valuable when evaluating new technology. A screen, an adaptive sequence and instant feedback may change the delivery system; they do not remove the need to define what a learner can do afterwards. Nor does an old proposal automatically settle whether a particular modern system works. Each system needs evidence matched to its learners and claims.

Previous chapter · Contents · Next chapter

Piaget’s developmental work adds another question: how does the organisation of thinking change, rather than merely how many correct responses accumulate? His discussion of developmental psychology collected in 1964, available through the Jean Piaget Foundation, considers maturation, experience, social transmission and equilibration. The last term concerns processes through which cognitive organisation is coordinated and reorganised; it is not simply a synonym for feeling calm.

For historical reading, the important distinction is between learning a particular answer and developing ways of organising relations. This does not justify reading a rigid timetable from a single classroom response. An observed error belongs first to a task, a setting and a learner’s explanation; assigning it to a theoretical stage requires a much stronger argument.

Original case: the arrangement changes, the count does not

Place six counters in a close row and six in a more widely spaced row. Ask a learner which row contains more counters, then ask how the answer was decided. If the longer row is selected, rearrange the counters while keeping their number visible and invite the learner to match them one to one.

This demonstration can reveal whether the learner’s answer is controlled by row length, counting, pairing or another interpretation of the question. It does not license a conclusion about the learner’s general intelligence or a comprehensive developmental classification. Language, familiarity with the task and the meaning assigned to “more” must remain under examination.

A useful repair is to make the invariant explicit: no counter was added or removed. Ask the learner to predict the result before another rearrangement and explain it afterwards. Then change the objects and the spacing. The goal is an independently usable relation, not successful imitation of the teacher’s last answer.

Developmental history thus widens the interpretation of an error. A wrong response can be evidence about the structure of a learner’s reasoning, but it is not self-interpreting evidence. Contemporary classroom decisions should use appropriate current research and professional judgment rather than treating a historical theory as an automatic placement instrument.

Previous chapter · Contents · Next chapter

Sweller’s 1988 paper, Cognitive Load During Problem Solving: Effects on Learning, examined a distinction between reaching a problem’s solution and acquiring knowledge of its structure. The paper proposed that conventional means–ends problem solving could consume processing capacity that would otherwise support schema acquisition. This is a specific theoretical and experimental argument, not a claim that thinking should be removed from education.

The historical change in emphasis is useful: effort is no longer treated as one undifferentiated good. A learner can expend effort searching for a route, remembering instructions, translating notation or understanding a relationship. These activities may have different consequences for the learning objective.

Worked algebra diagnosis

Take 3(x + 2) = 21. A complete solution divides both sides by 3 to obtain x + 2 = 7, then subtracts 2 to obtain x = 5. Substitution checks the result: 3(5 + 2) = 21. Another valid route expands first to 3x + 6 = 21, then subtracts 6 and divides by 3.

Suppose a learner copies either solution correctly but cannot explain why an operation must preserve equality. More copied examples may increase familiarity without addressing that missing relation. Ask the learner to identify what each line asserts and why the next line has the same solution. Then offer 4(x + 3) = 28 and request a justified first step before the full calculation.

If the learner understands equality but loses an intermediate value, leaving written steps visible may address a different difficulty. If the learner cannot multiply accurately, prerequisite arithmetic may need attention. These are alternative hypotheses to inspect, not diagnoses established by one equation.

The specialist article on working memory, cognitive load and mathematics lesson design develops this instructional route. Historically, the important contribution is a sharper question about what a task requires the learner to process. A difficult lesson is not automatically a productive one; an easier-looking lesson is not automatically superficial. Examine what understanding the design makes possible and what independent performance follows.

Previous chapter · Contents · Next chapter

Across these historical approaches, investigators and teachers need observable evidence. But a recorded score is not the psychological process itself. Counting correct answers can help compare performances; it does not eliminate the need to specify what the questions measure, who can access them and which interpretation the score can support.

A historian should therefore preserve the original task instructions and scoring rules whenever possible. “Accuracy improved” is incomplete if accuracy meant something different across occasions. A test with repeated familiar items may answer a question about retention of those items. It may not answer a question about selecting a method in a new context.

Worked case: equal totals conceal different needs

Imagine a ten-point classroom assessment with five points for calculating fractions and five for explaining equivalence. Learner A scores 5 + 1 = 6. Learner B scores 2 + 4 = 6. Both have 60%, but their profiles suggest different questions for follow-up. A may execute procedures without explaining the relation; B may understand the relation while making calculation errors. These are provisional interpretations to check through examples and discussion.

If the next lesson provides identical procedural drilling to both learners solely because their totals match, it may fail to address A’s difficulty. If it provides only verbal explanation, it may fail to address B’s. The same numerical score can support a broad statement about this assessment while being too coarse for a particular teaching decision.

Historical records create another complication. An assessment may have served selection, diagnosis, certification or programme evaluation. Those uses are not interchangeable. A threshold chosen to ration scarce places does not automatically represent a psychological boundary. A group average does not describe every learner in the group.

Ask whose work was counted, who was absent, which language was required and what accommodations were available. Missing information does not justify inventing an answer. It marks the limit of the conclusion. Careful measurement is not an obstacle to educational judgment; it is one of the ways judgment becomes accountable to the evidence it actually has.

Previous chapter · Contents · Next chapter

The 2007 What Works Clearinghouse guide Organizing Instruction and Study to Improve Student Learning illustrates a later form of research-to-practice work: an explicit set of recommendations with differentiated evidence ratings. It addresses distributed study, worked examples, representations, quizzing and explanatory questions. Its entries do not all carry the same rating, so citing the guide as one uniform guarantee misrepresents its structure.

More recent synthesis shows why the subject and comparison matter. Murray, Horner and Göbel’s 2025 mathematics meta-analysis found an overall spacing advantage of g = 0.28 across 27 studies. Its smaller set comparing testing with restudy produced g = 0.18 with a confidence interval crossing zero. That result limits a claim of a consistent mathematics testing effect; it does not establish that retrieval is useless.

A classroom meta-analysis published on 29 October 2026 examined 53 studies with 7,303 participants. It reported average effects of g = 0.37 for retrieval, 0.20 for distributed practice and 0.40 for interleaving, alongside moderator patterns and limitations concerning the evidence base. Different inclusion criteria and learning outcomes mean these estimates should not be treated as competing measurements of one identical intervention.

How to interpret the numbers

The symbol g here denotes a standardised effect-size estimate. It is not a percentage-point improvement, a percentage of pupils helped or a prediction that a particular child will gain that many marks. An effect of 0.28 cannot be translated into “28% better” without changing its meaning.

Before using a recommendation, write down the comparison. Retrieval rather than rereading is a different contrast from retrieval rather than a rich explanatory activity. Distributed rather than massed practice is different from one spacing schedule rather than another. The alternative matters because an instructional effect is estimated relative to something.

Then identify the intended outcome and timing. A next-day factual quiz and a later unfamiliar problem answer different questions. The mathematics retrieval-practice specialist offers a route into these distinctions. The historical advance is not the arrival of a final universal method. It is the increasingly explicit connection between a recommendation, a body of evidence and its boundaries.

Previous chapter · Contents · Next chapter

This original exercise brings the historical questions together. Assume two classes study proportional reasoning for the same scheduled time. Class A uses a new lesson sequence; Class B uses its usual sequence. Each class is assessed on a 20-point test before and after the lessons. The figures below are invented class means, and no random assignment is assumed.

Class A rises from 8 to 15, a gain of 7 points. Class B rises from 9 to 13, a gain of 4 points. The difference between gains is 7 − 4 = 3 points, equivalent to 15 percentage points on this 20-point scale. The final mean difference alone is 15 − 13 = 2 points. Those calculations answer different questions and should not be swapped.

What does the result establish?

It establishes the observed score changes under the stated measurement conditions. It does not establish that the new sequence caused the additional gain. The classes may differ in teaching, attendance, prior preparation or other experiences. The before-test helps describe starting performance but does not remove every possible difference. Without individual data, uncertainty estimates and a defensible design, a causal claim would outrun the information.

Now change the setting. A week later, pupils must compare two drink recipes rather than solve the familiar ratio format. Recipe P has 3 parts concentrate to 9 parts water. Recipe Q has 4 parts concentrate to 8 parts water. Which has the greater concentrate fraction, assuming volumes add and the ingredients are otherwise comparable?

For P, concentrate is 3 ÷ (3 + 9) = 1/4 = 25% of the mixture. For Q, it is 4 ÷ (4 + 8) = 1/3, approximately 33.3%. Q has the greater concentrate fraction. Comparing 3/9 with 4/8 uses concentrate-to-water ratios; those ratios can also order these mixtures, but they are not the concentrate fractions of the totals. The explanation must name the denominator.

Diagnose and repair the transfer failure

If pupils answer a familiar ratio item correctly but use the wrong denominator here, first ask them to identify the whole. A diagram with labelled concentrate and water can expose the relationship. Next remove the diagram and change the quantities. Finally ask for an explanation of why the denominator includes both ingredients.

Success after this support shows that the learner can complete the task with that intervention; it is not yet evidence of independent later transfer. Recheck with another suitable problem after a delay. James’s concern with teaching judgment, the early transfer question, Dewey’s inquiry and later attention to instructional support converge in the design of a better test—not in a claim that their theories are identical.

Previous chapter · Contents · Next chapter

A useful small history project begins with one claim, not an entire century. Choose a sentence such as “this exercise trains observation” or “this device lets every learner progress independently.” Locate the original publication, record its author and date, and identify whether the passage reports evidence, proposes a design or argues for an educational ideal.

Read enough surrounding material to recover the conditions. A sentence about a particular exercise can become misleading when detached from its task or population. In a digitised source, distinguish the original publication date, the date of the edition being scanned and the date the digital file was released. James’s linked volume is a practical example of why that matters.

A four-part research notebook

First, record the claim in your own words and preserve a precise location in the source. Second, describe the task and evidence. Third, state what the evidence does not establish. Fourth, propose a comparison or observation that could resolve one remaining uncertainty. Keep your proposed investigation separate from the historical author’s actual work.

For example, a report that students enjoyed a teaching device supports a statement about reported experience under those conditions. To examine independent learning, propose an assessment without the device. To examine whether it outperforms an alternative, specify the alternative, allocation method, time on task and outcome. One historical report can generate several research questions without answering all of them.

Look beyond celebrated authors where records permit. A preface may acknowledge teachers, collaborators or institutions that shaped a project. Dewey’s 1910 preface, for instance, credits his wife and colleagues connected with the Chicago Laboratory School. These acknowledgments help challenge a history told as if ideas emerged fully formed from isolated individuals.

Keep the project ethically modest. Use published records or voluntary adult learning demonstrations; do not create high-stakes classroom experiments or attach psychological labels to children for a history exercise. A good result is a more accurate interpretation of a source and a better-designed question. For a related institutional comparison, continue to The Gold Standard of History of Occupational Psychology, where the setting changes from classrooms to work.

Previous chapter · Contents · Next chapter

Who founded educational psychology?

No single name adequately represents all its questions, methods and institutions. James, Thorndike, Dewey and later researchers contributed different kinds of work. Calling someone a founder can be useful shorthand only if the account specifies what they helped establish. It should not erase earlier teaching traditions, collaborators or developments outside the sources selected here.

Is the history just a contest between behaviourism and cognition?

That contrast captures some theoretical disagreements but is too narrow for the whole subject. The sources also ask about educational purposes, developmental change, meaning, teacher judgment, measurement and transfer. A useful comparison identifies the learning problem and evidence each approach addresses before arranging them into opposing camps.

Does a historical text count as evidence that a method works today?

It is evidence of what its author wrote, proposed or reported. Whether the method works for a present group requires evidence matched to that claim. A historical experiment may remain informative, but its task, participants, comparison and outcome must be examined. Age alone neither validates nor refutes an argument.

Should every lesson maximise retrieval or minimise cognitive load?

Neither instruction is precise enough to guide every lesson. Ask what is being learned, what the learner already knows, what the alternative activity would be and what independent outcome is intended. A practice method can be useful without being suitable for every phase or every type of learning. The cases above show how changing the task changes the decision.

How can a learner use this history without becoming a specialist?

Start with one everyday difficulty: remembering a rule, explaining a definition or solving a changed problem. Read the chapter that sharpens the question, try its worked example, then explain which observation would distinguish understanding from successful imitation. Keep a short record of assumptions and revise it when new evidence appears.

Which sources should I read first?

Begin with James’s opening discussion of psychology and teaching, then compare the purpose of Thorndike and Woodworth’s transfer investigation with Dewey’s account of reflective thought. Read Skinner’s proposed mechanism alongside his evaluation qualifications. Use Piaget and Sweller to ask different questions about development and instructional demands. Finally compare the 2007 guide with the two recent meta-analyses, paying attention to their dates, comparisons and scope. The primary and official links appear beside the claims they support throughout this guide.

For the next stage of study, move from this historical account to Educational Psychology for present-day mechanisms, then select a concrete explanation from How X Works. Return to Education: Learning Transfers into Life to connect the lesson with its wider purpose, or browse The Gold Standard master for another subject route.

Previous chapter · Contents · Gold Standard master