A learner solves ten problems correctly in class. The eleventh problem changes the wording, removes the familiar diagram, and places the same idea inside a new context. The learner stops.
The knowledge may have been learned. The transfer has not yet been demonstrated.
Transfer is one of the most demanding tests of teaching because it asks whether learning survives when the surface changes. The learner must recognise what matters, ignore what does not, select relevant knowledge, reconstruct the route, and apply it without the original lesson announcing which idea is needed.
Transfer works in teaching when the teacher makes the deep structure of a skill or concept visible, varies the surface features around that structure, teaches learners to discriminate between similar and different cases, removes contextual cues gradually, and verifies that learners can select and reconstruct the relevant knowledge in genuinely changed conditions.
This article continues the eduKate Sengkang How Teaching Works series after How Scaffolding and Fading Work in Teaching. It owns the general teacher-side architecture of transfer.
It does not replace the learner-side How Learning Transfer Works | When Knowledge Survives a New Problem, How to Improve Learning Transfer, subject-specific transfer guides, or Tutor Handbook owners such as the Example-Variation Gate and representation-translation controls. Those pages retain their learner, repair, subject and tutoring jobs.
The worked examples below are original teaching examples unless an external source is named. They are not records of actual students or official examination questions.
A route through transfer in teaching
Begin with what transfer is, then examine deep structure, surface variation, discrimination, near and far transfer, representation changes, method selection, scaffold removal, and verification. Worked cases cover Mathematics, Science, English and writing, followed by whole-class, small-group, examination, homework and AI applications.
Transfer is not repetition with different numbers
If a learner completes a second problem that differs only by one number, that is useful practice. It is weak evidence of broad transfer.
Transfer requires some meaningful change: wording, representation, context, method cue, task order, surface story, data form, combination with another skill, or the absence of a scaffold that was previously present.
The learner must still recognise the underlying relationship despite that change.
Transfer is a selection problem as much as an execution problem
A learner can know how to execute a method and still fail to transfer because the learner does not know when the method applies.
This is why mixed and changed problems matter. They make selection part of the performance.
Teacher-side transfer therefore includes two jobs: preserve the procedure or concept, and teach the learner to recognise its conditions of use.
Transfer is not guaranteed by understanding one example deeply
Deep understanding helps, but transfer still requires experience with variation and selection.
A learner may understand one fraction model thoroughly and still fail when the same relation appears on a number line. A reader may understand inference in one story and fail when the next passage presents evidence through dialogue rather than description.
Teachers should therefore build a route from understanding to changed conditions.
Transfer begins by deciding what should travel
Before designing varied practice, name the invariant.
- What principle should remain true?
- What relationship should be recognised?
- What decision rule should survive?
- What representation should be reconstructable?
- What checking routine should reappear?
- What surface details are allowed to change?
If the teacher cannot state what should travel, variation can become random novelty rather than transfer design.
Teach the deep structure explicitly
Experts often see structure that novices see only as surface detail.
A mathematician sees a proportional relationship beneath a shopping story. A scientist sees conservation beneath a new apparatus. A strong reader sees evidence-to-claim structure beneath a different narrative. A writer sees purpose-audience-effect beneath a different genre.
Teachers should make this structure visible rather than assume repeated exposure will reveal it automatically.
Name the invariant after the example, not instead of the example
Abstract rules can be empty if learners have no concrete cases to attach them to.
Use examples first or alongside the principle, then ask what stayed the same across them.
For example, after comparing several ratio problems, ask what makes them proportional. After several inference passages, ask what all strong answers do with evidence.
Use multiple examples that share structure but differ in surface
One example can accidentally teach irrelevant features. Multiple deliberately varied examples help learners see what changes and what does not.
The IES / WWC Organizing Instruction and Study to Improve Student Learning guide notes that variability across worked examples and problems can increase study demands but improve later learning. The mathematics problem-solving guidance also emphasises comparing worked examples and representing problems in ways that make structure visible.
Ask learners to compare examples explicitly
Do not rely on learners to notice the common structure spontaneously.
Ask: “What is the same?” “What changed?” “Which feature determines the method?” “Which details are irrelevant?” “Why do these two different-looking problems use the same idea?”
Comparison turns variation into a learning event.
Vary surface features systematically
Useful surface variation can include numbers, names, contexts, object types, diagram layouts, wording, order of information, irrelevant information, units, representations and question formats.
Change one or two features early. Increase variation later.
If everything changes at once, failure becomes difficult to interpret.
Surface variation should not accidentally change the target skill
A transfer task can become a different task if it introduces new vocabulary, unfamiliar mathematics, a different representation and a new reasoning demand simultaneously.
Teachers should know which feature is being varied and which capability is being held constant.
This is the teacher-side transfer version of the Tutor Handbook’s Example-Variation Gate.
Use irrelevant information deliberately after the core structure is stable
Real problems contain details that do not determine the solution.
Once the learner understands the core relation, add information that must be ignored. This tests whether the learner can identify structure rather than copy every visible quantity into a calculation.
Change the order of information
Learners can become dependent on the sequence used in worked examples.
Present the same information in a different order. Ask the learner to reconstruct the representation before solving.
Change the story while keeping the relationship
A ratio can appear in recipes, maps, speed, mixture, scale drawings and finance. An inference can appear in narrative, dialogue, informational text and visual sources.
Context variation helps learners see the relation as portable rather than tied to one story.
Transfer depends on discrimination between similar and different cases
Learners must not only see sameness. They must also know when two similar-looking tasks require different methods.
This is a crucial but often neglected transfer skill.
Use contrast cases to teach method boundaries
Place two similar problems side by side where one is proportional and the other is not. Compare a correlation claim with a causal claim. Compare an inference question with an explicit-detail question. Compare two writing prompts that require different audience choices.
Ask what feature changes the route.
Teach non-examples as carefully as examples
A non-example helps define a concept boundary.
After teaching direct proportion, show a relationship that is linear but not proportional. After teaching evidence-based inference, show an interpretation that is possible but unsupported.
Transfer improves when learners know what the concept is not.
Mixed practice makes discrimination unavoidable
Blocked practice says, “Use this method now.” Mixed practice asks, “Which method is appropriate?”
Use mixed practice after the individual methods are stable enough to discriminate.
Mixing too early can produce noise rather than useful transfer practice.
Near transfer and farther transfer should be sequenced
Near transfer changes relatively little. The learner uses the same principle in a slightly different example.
Farther transfer changes more: context, representation, task structure or combination with other knowledge.
Teachers should not jump immediately from one modelled example to a distant application and conclude that failure means the concept was never learned.
Build a transfer gradient
- same structure, different numbers;
- same structure, different wording;
- same structure, different representation;
- same structure, different context;
- same structure mixed with a neighbouring alternative;
- same structure embedded inside a larger task;
- same structure after delay;
- same structure under realistic performance conditions.
The gradient makes failure more interpretable because the teacher can see where transfer breaks.
A failed far-transfer task should not automatically restart teaching from zero
Find the first transfer distance that succeeds.
If the learner handles changed numbers and wording but fails when representation changes, representation translation is the likely teaching target. If the learner succeeds until methods are mixed, discrimination is the target.
Repair the missing bridge, not the entire original concept.
Transfer across representations should be taught explicitly
Knowledge can become trapped inside one representation.
A learner can understand a fraction strip but not a number line, a graph but not an equation, a paragraph frame but not an outline, a causal diagram but not prose.
Teach learners to translate between forms and explain what remains invariant.
Concrete, diagrammatic, verbal and symbolic forms should connect rather than compete
One representation may make a relation visible; another may be more efficient for advanced work.
Move between them deliberately. Ask what each shows well and what it hides.
Remove representation labels once translation is stable
Eventually the learner should decide whether a table, graph, equation, diagram or prose explanation is the best representation.
That decision is part of independent transfer.
Transfer requires method selection under uncertainty
In authentic work, the method is rarely named in the margin.
Teaching should therefore move from method-labelled practice to structure-labelled practice to unlabelled mixed practice.
Ask learners why one method does not fit
Selection becomes stronger when learners can reject a tempting alternative.
“Why is this not a percentage-change problem?” “Why does this evidence not justify causation?” “Why is this not an inference question?”
Rejecting a wrong method reveals boundary knowledge.
Use self-explanation to make selection criteria explicit
Ask learners to explain what feature triggered the method.
Later, shorten the explanation requirement as selection becomes fluent.
Transfer requires contextual cues to fade
Chapter headings, colour coding, worksheet sections, example order and teacher language can all become hidden scaffolds.
Remove them gradually when future performance will not provide them.
This connects directly to How Scaffolding and Fading Work in Teaching.
Transfer should be tested after delay as well as surface change
A fresh-looking problem one minute after the lesson can still benefit from highly active memory.
Return later. Combine changed surface with delayed access.
That creates stronger evidence that knowledge has become durable and portable.
Transfer needs retrieval
The relevant knowledge must return before it can be applied.
Retrieve the principle, then select it in context. Over time, the retrieval should occur inside the application without a separate warm-up announcing it.
See How Retrieval Practice Works in Teaching.
Transfer needs spacing
Knowledge that has only ever been used immediately after teaching may fail when the curriculum returns to it later.
Use spaced transfer tasks so the learner must both retrieve and apply.
See How Spacing Works in Teaching.
Transfer needs feedback on the selection, not only the answer
If the learner chooses the wrong method, feedback should address why that method seemed appropriate and which feature should have triggered another choice.
A correct final answer produced by an inefficient or accidental route may also deserve discussion when strategic transfer is the target.
Worked examples should become transfer examples
After learners study one worked solution, show another with different surface details and ask which decisions remain the same.
Later, give the changed problem without the example.
Counterexamples prevent false transfer
Learners can overgeneralise a recently taught method and use it everywhere.
Include cases where the method should not be used. Ask learners to identify the stopping condition.
Transfer can fail because the learner retrieves the wrong level of abstraction
The learner may remember “multiply by 100” instead of the percentage relationship, “quote evidence” instead of evidence-to-claim reasoning, or “use a diagram” instead of the underlying geometry.
Teach the relationship at the right level of abstraction so it can travel.
Transfer can fail because the original example was too narrow
If every early example shares irrelevant surface features, learners may encode those features as part of the concept.
Use deliberate early variation once the core relation is understandable.
Transfer can fail because the learner lacks background knowledge in the new context
A learner may know the mathematics but not understand the financial vocabulary in the transfer problem. A reader may know inference but lack the cultural or lexical knowledge needed to interpret the passage.
Distinguish failure of transfer from failure of access to the new context.
Transfer can fail because working memory is overloaded
The learner may possess the relevant knowledge but lose it when several new demands combine.
Simplify the transfer context enough to identify the first failing bridge, then rebuild complexity.
Transfer can fail because the teacher has always announced the method
A learner who solves “quadratic equations” worksheets may not recognise a quadratic relation in a modelling task.
Remove method labels before high-stakes performance.
Transfer can fail because feedback repaired the product, not the decision rule
If a learner only learns how to fix one answer, the same error can appear in a different form.
Feedback should identify the transferable principle whenever possible.
Transfer can fail because practice is too predictable
Fixed order, identical wording and chapter-labelled sets create strong contextual support.
Introduce unpredictability gradually enough that the learner still knows what is being learned.
Transfer can fail because practice is too variable too soon
Novices may need stable early examples to form the first schema.
Variation should grow with knowledge.
Transfer can fail because the new problem requires a second untaught skill
A far-transfer task can inadvertently add difficult reading, complex arithmetic or unfamiliar tools.
Control those added demands when diagnosis matters.
Transfer can fail because the learner does not expect transfer
Students often compartmentalise subjects and chapters. They may not search memory beyond the current unit.
Teach a habit of asking: “What have I seen before that has the same structure?”
Transfer can be supported by analogical comparison
Place two different-looking problems side by side and ask learners to map corresponding elements.
The mapping should identify deep relations, not merely similar nouns or numbers.
Transfer can be supported by learner-generated examples
Ask learners to invent a new problem that uses the same principle but looks different.
Generating a valid transfer case requires understanding of the invariant and its boundary.
Transfer can be supported by learner-generated non-examples
Ask learners to design a tempting case where the method would be wrong.
This tests discrimination and concept boundaries.
Transfer can be supported by “same or different?” routines
Give two tasks and ask whether the same underlying idea applies. Require a reason.
The routine is simple enough for Primary learners and deep enough for advanced subjects.
Transfer can be supported by changed-question practice
Keep the underlying information but change what is being asked.
This prevents learners from associating one dataset or passage with one fixed method.
Transfer can be supported by changed-information practice
Keep the question type but change which information is relevant or irrelevant.
The learner has to reconstruct the evidence path.
Transfer can be supported by changed modality
Move from prose to graph, diagram to equation, oral explanation to written explanation, data table to narrative case.
Ask what relation remains constant.
Transfer can be supported by delayed problem revisiting
Bring back the same principle weeks later inside a new unit.
Do not announce the connection immediately. Let learners search for it, then teach the bridge if needed.
Transfer should be visible in curriculum design
A curriculum that never reuses old knowledge in new contexts cannot reveal whether transfer is developing.
Map where important ideas should reappear across the year.
Transfer should be designed across subjects where the relationship is genuinely shared
Evidence-to-claim reasoning appears in Science, English and Humanities, but the evidence rules differ by discipline.
Teach the shared structure and the subject-specific boundary.
Do not force false cross-subject transfer
Similar vocabulary can hide different meanings. “Evidence,” “function,” “model,” “power” and “significant” change meaning across subjects.
Cross-subject transfer must respect disciplinary definitions.
Primary transfer should begin with visible contrasts
Younger learners benefit from concrete changes they can compare: different numbers, pictures, contexts and wording while the same relationship remains visible.
Ask learners to name what stayed the same.
Secondary transfer should increasingly require method selection
Secondary learners face more neighbouring concepts and methods.
Mixed sets, unfamiliar contexts and representation changes become important because examinations rarely preserve chapter labels.
JC and advanced transfer should include assumptions and limitations
Advanced transfer is not merely applying a formula elsewhere. Learners must know which assumptions still hold, what changes under a new model, and when the original method is no longer appropriate.
Worked case: Mathematics transfer from direct proportion to changed context
Initial example: 3 notebooks cost $12, so 1 notebook costs $4 and 5 notebooks cost $20.
Near transfer: 7 identical pens cost $21. Find the cost of 11 pens. The learner identifies the same proportional structure.
Representation transfer: provide a table of quantity and total cost with one row missing.
Discrimination case: a taxi fare includes a fixed starting charge plus a per-kilometre charge. The relation is not directly proportional because the graph does not pass through the origin.
Farther application: compare exchange-rate conversion, recipe scaling and scale drawing. Ask which are proportional and what invariant ratio controls each.
The transfer target is not “use unitary method whenever two numbers appear.” It is recognise multiplicative structure and its boundary.
Worked case: Science transfer from insulation model to unfamiliar thermal system
Initial learning: insulation reduces the rate of energy transfer; it does not stop transfer completely.
Near transfer: compare a covered and uncovered cup.
Context transfer: explain why a jacket reduces energy transfer from the body to colder surroundings.
Boundary case: ask why the same jacket does not actively generate thermal energy under ordinary use.
Representation transfer: interpret a temperature-time graph for two containers without being told which line represents insulation.
Farther transfer: analyse a cool box or building insulation example and identify which energy-transfer pathways matter.
Worked case: English transfer from inference routine to unfamiliar text
Initial passage: a character hides a damaged object and changes the subject when asked about it.
Near transfer: a new narrative with a different character and different behavioural cues.
Representation transfer: infer attitude from a text-message exchange rather than narrative prose.
Boundary case: provide a passage where the text supports uncertainty rather than one strong motive.
Farther transfer: use an informational text and ask what conclusion the evidence supports without overclaiming.
The invariant is evidence-constrained reasoning, not a fixed sentence frame.
Worked case: writing transfer from analytical paragraph to new genre
Initial structure: claim, evidence, explanation, relation to purpose.
Near transfer: a second literary-analysis paragraph with different evidence.
Context transfer: a persuasive letter where evidence supports a recommendation.
Farther transfer: a Science evaluation where evidence must support a methodological judgement.
The teacher should discuss what remains common—claim-evidence relation—and what changes by discipline, audience and genre.
Small-group teaching can make transfer breakdown visible
In a three-student group, each learner can receive the same transfer task after individual first attempts.
One learner may fail at retrieval, another at representation translation, and another at method selection. The teacher can then vary the support precisely.
Group discussion should follow enough individual work to preserve diagnostic evidence.
Whole-class transfer requires broad response evidence
Do not infer class transfer from a few confident volunteers.
Use short simultaneous responses, mini-whiteboards, classification tasks or brief written explanations where appropriate.
Homework can be designed as a transfer environment
Homework removes teacher presence and can change context naturally.
Give one familiar application, one changed context, and one mixed item where the method is not named.
Record where external help was needed so transfer evidence remains interpretable.
Examination preparation should test transfer before full papers
Move from chapter-labelled sets to mixed sets, then unfamiliar wording, then timed sections and full papers.
Full papers are rich transfer environments but diagnostically noisy. Use targeted changed problems to repair specific transfer gaps between papers.
AI can generate transfer variants and also fake transfer
AI can create changed contexts, representations and distractors quickly.
But if AI also identifies the method or explains the mapping before the learner attempts it, the transfer decision has been outsourced.
Use AI to generate candidate variants, verify them, let the learner attempt independently, and request hints only after the transfer state becomes visible.
Transfer prompts should fade
Early prompt: “Which previous example has the same structure?”
Later prompt: “What do you recognise?”
Final condition: no prompt. The learner initiates the search for prior knowledge.
A transfer record should describe the changed condition
A useful note is not simply “transfer successful.”
Target structure → original context → changed feature → support → result → next variation.
Example: “Direct proportion → shopping → graph representation → no cue → selected correctly → next mix with affine/non-proportional cases.”
Transfer should have an evidence ladder
- same problem structure, new numbers;
- new wording;
- new representation;
- new context;
- mixed neighbouring cases;
- embedded larger task;
- delayed return;
- realistic independent conditions.
The ladder is not a universal sequence. It is a way to make transfer claims proportionate to evidence.
Verify transfer without leaking the answer
A transfer test should not name the method if method selection is part of the claim.
It should not keep the model answer visible if reconstruction is part of the claim.
It should not allow the strongest group member to answer first if individual transfer is the claim.
It should resemble the authentic support conditions of the future task.
Use delay to separate immediate adaptation from durable transfer
After a successful changed problem, return again later.
The second changed problem should not be an exact copy of the first transfer test.
Use mixed contexts to test selection
Present several problems where only some require the target principle.
The learner must identify which ones and explain why.
Use changed representations to test portability
Move between graph, table, prose, diagram and symbols where the subject permits.
Ask the learner to reconstruct the invariant relation.
Use self-explanation to inspect transfer reasoning
A correct answer alone may come from superficial similarity.
Ask why the old knowledge applies here. The explanation can reveal whether the learner recognised deep structure or guessed from surface resemblance.
Use counter-transfer cases to prevent overgeneralisation
After learners succeed on several transfer cases, include a similar-looking problem where the target method is wrong.
This tests whether transfer has become indiscriminate method use.
A practical teacher sequence for transfer
- Define the invariant that should travel.
- Teach the invariant through clear examples and explanation.
- Check the learner can use it in the original context.
- Vary one surface feature.
- Ask what stayed the same.
- Introduce a contrasting non-example.
- Change representation.
- Remove method labels.
- Mix neighbouring cases.
- Ask learners to justify selection.
- Change context.
- Embed the knowledge inside a larger task.
- Fade transfer prompts.
- Return after delay.
- Test under authentic support conditions.
- Record the first transfer distance that fails.
- Repair that bridge rather than reteach everything.
- Repeat with new variation.
- Retire special transfer support when selection becomes reliable.
Common transfer failure modes
- same worksheet with different numbers is called transfer;
- the invariant is never made explicit;
- surface features change so much that a second skill becomes the real barrier;
- method labels remain visible forever;
- blocked practice is never replaced by mixed selection;
- variation is introduced before the original schema is stable;
- learners compare examples without discussing what is structurally common;
- non-examples are omitted, so boundaries remain unclear;
- one successful changed problem is treated as broad transfer;
- no delayed transfer check occurs;
- the learner receives transfer prompts every time;
- teacher feedback fixes the answer but not the selection rule;
- background knowledge in the new context is ignored;
- group success is treated as individual transfer;
- AI identifies the mapping before the learner does;
- far-transfer failure causes complete reteaching rather than bridge diagnosis.
Evidence, interpretation and limits
Transfer is difficult to study because “new problem” can mean many different distances from the original learning. Evidence from instructional research supports several mechanisms used in this article—worked-example comparison, varied practice, delayed review, multiple representations, deep explanatory questioning and problem-structure teaching—but no single technique guarantees far transfer.
The IES / WWC Organizing Instruction and Study to Improve Student Learning guide recommends varying worked examples and problems after instruction and integrating abstract and concrete representations. The Improving Mathematical Problem Solving in Grades 4 Through 8 practice guide supports teaching visual representations, monitoring the problem-solving process and comparing multiple solution strategies. These are subject-specific and older guidance sources, so they should not be treated as a universal transfer recipe.
AERO’s current Vary Practice guidance places varied and spaced practice inside a broader teaching model. It is useful for implementation principles but does not imply that every change of context produces transfer automatically.
The original transfer gradients, worked cases and teacher routines in this article are editorial teaching designs. They have not been evaluated together as one intervention. No universal amount of variation, fixed near-to-far sequence or guaranteed effect size is claimed.
Sources for this edition were reviewed on 22 September 2026.
The transfer standard: the learner should recognise the idea when the lesson stops naming it
Transfer is not proven because the learner can repeat the method in the same chapter.
The stronger outcome is that the learner recognises deep structure beneath changed surface features, rejects tempting wrong methods, reconstructs the relevant representation, uses the knowledge after delay, and does so when the new task does not announce which old lesson it came from.
Teaching has transferred when the learner can find the old knowledge inside a genuinely new problem and make it useful again.
Continue through How Teaching Works, use the learner-side How Learning Transfer Works, or revisit How Scaffolding and Fading Work in Teaching.
