The 50-Second Read
Transfer is not recognising that two questions look alike. It is recognising that the same underlying relationship still holds after the surface has changed.
A familiar problem may change its numbers, wording, diagram, context, representation, command word, direction of reasoning or the order in which information appears. If the deep structure remains, an old method may transfer. If the structure changes, copying the old method can become the trap.
The operating loop is:
Identify the target → strip away surface detail → name the invariant relationship → map the new givens onto the old structure → recheck the method’s conditions → adapt what changed → execute → verify that the transferred method still fits the new problem.
This article is the next examination-performance edge in the How to Think Properly series. Know When a Familiar Pattern Is Misleading owns the danger signal: deciding whether familiarity is false permission. How Intelligence Works | Transfer and Recomposition owns the broad cognitive architecture of how knowledge survives changed problems. How Intelligence Works | Abstraction owns the general operation of removing surface detail and keeping structure. How Intelligence Works | Analogy owns structure crossing between domains. This page owns a narrower practical job: how a student transfers an already-known examination method to a changed question while preserving its conditions of validity and adapting only what the new surface genuinely changes.
One-Sentence Definition
Examination transfer is the controlled reuse of a known reasoning structure in a changed problem after verifying that the invariant relationships and method conditions still hold.
The Question That Looks New Only Because the Surface Moved
Clara opens a Mathematics paper and sees a context she has never practised.
A drone rises from a platform while a camera tracks its horizontal displacement.
Nothing about the problem looks like the school examples.
No ladder.
No building.
No familiar triangle diagram.
Clara says, “We never learnt this.”
Jo asks three questions.
- What quantity is the question asking for?
- What geometric relationship exists among the given quantities?
- What changed from the examples you know—and what did not?
Clara redraws the motion as a right triangle.
The story changes.
The structure does not.
A few minutes later, Ethan encounters the opposite problem.
His question looks almost identical to one he practised.
He applies the old method immediately.
One condition has changed.
The old method no longer applies.
Clara failed to transfer because the surface looked too different.
Ethan transferred when he should not have because the surface looked too similar.
Good transfer lives between those failures.
Surface and Structure
Every examination problem has surface features and structural features.
Surface features include:
- names;
- numbers;
- story context;
- diagram appearance;
- order of information;
- vocabulary;
- format;
- whether data appear in prose, graph, table or formula;
- topic labels;
- familiar keywords.
Structural features include:
- which quantities depend on which;
- what remains invariant;
- what constraints must hold;
- what operation the question requires;
- what causal direction links variables;
- which evidence supports which claim;
- what mathematical object must be produced;
- which theorem conditions are satisfied;
- what decision criterion governs the judgement;
- what state transition an algorithm performs.
Transfer succeeds when the student can ignore enough surface variation to see preserved structure without ignoring so much detail that changed conditions disappear.
The Invariant Test
The fastest transfer question is:
What must remain true for the old method to still work?
That “must remain true” is the invariant or condition set that gives the method permission.
For a percentage increase method, the reference base must still be the original quantity.
For similar-triangle ratios, the similarity conditions must still hold.
For a causal Science explanation, the same mechanism must still connect the changed condition to the measured outcome.
For an English inference method, the answer still needs to move from textual evidence to implied meaning.
For a source evaluation, the judgement still has to be made relative to a specific claim or inquiry.
For an algorithm, the invariant and boundary conditions must still survive the altered input.
Transfer Is a Mapping Problem
Students sometimes describe transfer as “remembering the right method.”
That is incomplete.
The method is remembered.
Transfer happens when the learner maps the new problem onto the method’s structure.
A useful mapping table is:
- Old target → new target
- Old givens → new givens
- Old relationship → new relationship
- Old constraints → new constraints
- Old representation → new representation
- Old answer form → new answer form
If the mapping is coherent and the conditions survive, reuse becomes justified.
The Transfer Loop
- Strip: remove names, context and decorative details.
- Name: state the underlying relationship or job.
- Match: find the known structure that resembles it.
- Check: verify the old method’s conditions.
- Map: assign the new quantities, evidence or states to the old roles.
- Adapt: change what the new problem changes.
- Run: execute the transferred method.
- Verify: test fit against the new problem, not the remembered one.
The loop prevents both under-transfer and over-transfer.
Near Transfer
Near transfer occurs when the new task differs only modestly from practice.
Numbers change.
Names change.
Diagram orientation changes.
The question is worded differently.
The order of givens changes.
Near transfer is where fluency should begin to show.
Students should not need to rebuild the method from first principles every time.
But even near transfer requires condition checking when one small change can invalidate the route.
Farther Transfer
Farther transfer changes more of the surface and sometimes the domain.
A ratio idea moves from recipe scaling into map scale.
A conservation principle moves from one apparatus into another.
An evidence-inference relation learned in comprehension moves into source analysis.
A state-invariant idea moves from algebraic checking into algorithm tracing.
Farther transfer usually requires stronger abstraction because fewer surface cues announce the connection.
Do Not Confuse Transfer With Template Copying
A template can support transfer and can also destroy it.
If the student copies the sequence because the page looks familiar, that is imitation.
If the student knows what each step accomplishes, checks the conditions and changes the step when the new problem changes, that is transfer.
Template ownership is visible when the learner can answer:
- Why is this step here?
- What condition makes it valid?
- What would make me remove it?
- What would make me replace it?
- Which part is structural and which part was specific to the example?
The Representation Transfer Problem
Students often know a relationship in only one representation.
They know it as:
- a formula but not a graph;
- a graph but not a verbal rule;
- a diagram but not an equation;
- a paragraph structure but not an evidence map;
- a table but not a rate relationship;
- code but not an algorithmic invariant.
Examinations frequently test whether the learner can recognise the same structure after representation changes.
Strong transfer therefore includes translation.
The Direction-Reversal Problem
A familiar method may be learned forward and examined backward.
Practice gives cause and asks for effect.
The exam gives effect and asks for cause.
Practice gives a formula and asks for output.
The exam gives output and asks for a parameter.
Practice gives evidence and asks for inference.
The exam gives a claim and asks for evidence.
Transfer succeeds when the learner understands relationships bidirectionally rather than memorising one execution direction.
The Command-Change Problem
The topic can remain identical while the command changes from describe to explain, explain to evaluate, find to show, identify to justify, or compare to judge.
This is not a small wording change.
The underlying knowledge may transfer while the answer operation changes.
A strong learner transfers the content structure and rebuilds the output form around the new command.
The Scale-Change Problem
Some methods survive scale changes.
Some do not.
A physical model may work at one range and fail at another.
An approximation may be valid for small changes and fail for large ones.
A social example may illustrate a mechanism without supporting population-wide generalisation.
Transfer requires checking whether changing scale preserves the assumptions that made the old method work.
The Boundary-Change Problem
A method can work perfectly on ordinary cases and fail at boundaries.
Zero.
Empty input.
Equal values.
Maximum or minimum values.
A threshold exactly met.
An argument where both sources agree rather than disagree.
Boundary changes are useful transfer tests because they reveal whether the method depends on a hidden ordinary-case assumption.
Mathematics | Transfer the Relationship, Not the Story
Mathematics transfer often begins by stripping context.
A train problem, filling-tank problem, recipe problem and financial problem may all contain the same proportional structure.
A ladder, shadow, drone and line-of-sight problem may all reduce to right-triangle relationships.
A discount, growth rate and percentage change problem may all depend on selecting the correct reference base.
The transferable object is the relationship.
Mathematics Case | Ratio Across Contexts
A student has learned recipe scaling.
The exam asks about a map.
The nouns differ.
The invariant is multiplicative comparison between corresponding quantities.
Transfer succeeds when the student maps recipe quantity to map distance and scaling factor to scale ratio while preserving units and direction.
Mathematics Case | Quadratic Structure Without the Word “Quadratic”
A context produces a relationship that becomes quadratic after modelling.
No keyword announces the topic.
The student must recognise the structural form after representation, not from the surface language.
Transfer begins once the equation exists.
Then familiar solving methods become available—subject to domain and answer-form checks.
Mathematics Case | Same Diagram, Different Condition
A geometry question looks almost identical to a practised one, but one pair of lines is no longer parallel.
This is the danger side of transfer.
The old angle relationship loses permission.
Strong transfer means preserving the diagram-reading habits while rejecting the method whose condition disappeared.
Science | Transfer the Mechanism, Not the Memorised Example
Science transfer often depends on recognising a mechanism in a new setting.
A force principle appears in a different apparatus.
Diffusion appears in a biological rather than laboratory context.
Energy transfer appears in a new device.
Conservation appears with different measured quantities.
The learner should ask:
What mechanism is invariant even though the apparatus changed?
Science Case | Same Mechanism, Different Apparatus
A student learned that increasing temperature changes reaction rate using one standard laboratory setup.
The exam presents a completely different reaction and measurement method.
If the same particle-level mechanism remains relevant, the explanation can transfer.
The student must adapt the observed outcome and variables without copying the old apparatus language.
Science Case | Same Data Pattern, Different Claim Strength
Two graphs show similar correlations.
One comes from a controlled experiment.
The other is observational.
The surface pattern transfers.
The causal conclusion does not automatically transfer.
Method transfer must carry evidence conditions with it.
English | Transfer the Evidence Operation, Not the Sentence Frame
English students often learn sentence frames that work on one task and become brittle on another.
The transferable structure is usually deeper:
- evidence → inference;
- wording → effect;
- feature → purpose;
- Text A → shared dimension → Text B;
- claim → evidence → reasoning;
- proposition → criterion → judgement.
The wording of the paragraph may change.
The reasoning operation transfers.
English Case | New Genre, Same Inference Operation
The student practised inference on narrative fiction.
The exam uses a speech.
The genre changes.
The operation remains: identify evidence, infer implied attitude or purpose, and support the inference with text.
Transfer fails if the learner thinks inference belongs only to stories.
English Case | Same Technique, Different Effect
Metaphor appears in two passages.
The first creates vulnerability.
The second creates aggression.
The technique label transfers.
The effect does not.
Strong transfer preserves the operation “wording → contextual effect” while rebuilding the effect from the new wording.
Humanities | Transfer the Criterion and Causal Logic
Humanities transfer can become shallow when students memorise factor lists.
The deeper transferable structures include:
- background condition → trigger → event;
- factor → mechanism → consequence;
- source property → claim-specific limitation;
- evidence → criterion → importance judgement;
- change → continuity → extent;
- provenance → purpose → evidential consequence.
Facts change from topic to topic.
The reasoning architecture can transfer.
Humanities Case | New Event, Same Causal Architecture
A student learned to distinguish long-term cause and trigger in one historical topic.
A new topic uses different actors and events.
The factual content does not transfer.
The temporal-causal architecture does.
The learner asks which conditions created vulnerability and what event activated them.
Computing | Transfer the State Logic, Not the Code Shape
Programming transfer fails when students copy syntax without understanding state.
The transferable structures include:
- input → transformation → output;
- initial state → loop invariant → termination;
- condition → branch → state update;
- boundary → guard → safe behaviour;
- test case → expected state → diagnostic result.
Variable names and code layout can change completely while the algorithmic structure remains.
Computing Case | Same Algorithm, Different Data Type
A loop that finds a maximum in a list of numbers appears again with records containing scores.
The syntax and comparison target change.
The invariant remains: after each processed element, the stored candidate is the best seen so far under the chosen criterion.
Transfer succeeds when the learner carries the invariant and adapts the comparison field.
Computing Case | Same Surface, Different Boundary
A loop looks identical to practice except the input may now be empty.
The ordinary-case algorithm may no longer be safe.
The student must transfer the core logic while adding a guard.
This is adaptation, not blind reuse.
The Transfer Failure Taxonomy
- Surface lock: cannot recognise structure when wording changes.
- Template copying: repeats steps without checking roles.
- Condition loss: transfers method but forgets permission requirements.
- Over-abstraction: ignores a changed detail that actually matters.
- Under-abstraction: treats every context as entirely new.
- Direction lock: can use a relationship only in one direction.
- Representation lock: knows the structure in only one form.
- Command lock: transfers content but not answer operation.
- Boundary blindness: transfers ordinary-case logic to edge cases.
- Analogy overreach: maps too many features from the old case.
Ben | Transfer by Fast Structural Check
Ben transfers quickly and can over-transfer.
His routine is:
Same job? Same relationship? Same conditions? Then go.
The routine preserves his speed while protecting against false familiarity.
Aisha | Transfer by Reconstructing the Relationship
Aisha can lose access when familiar cues disappear.
Her transfer routine begins with:
What relationship do I know that could connect these givens to this target?
This starts from structure rather than a missing keyword.
Ryan | Transfer Without Endless Comparison
Ryan sees several possible analogies and distrusts all of them.
His rule is:
Find the closest structural match, test its conditions once, then commit unless a concrete contradiction appears.
Mira | Transfer Through External Mapping
Mira benefits from writing role mappings:
- old x → new quantity;
- old constraint → new constraint;
- old target → new target;
- old invariant → unchanged relationship.
This reduces working-memory load when several surface features change at once.
Clara | Transfer by Escaping Surface Lock
Clara’s first question is not “Have I seen this?”
It becomes:
What structure have I seen that behaves like this?
The change shifts retrieval from visual memory to relational memory.
Ethan | Transfer Without Analogy Overreach
Ethan sees deep analogies quickly and can carry too much across.
His check is:
Which mapped features are necessary to the method, and which are merely similarities?
Good analogy transfers structure selectively.
Training Drill 1 | Same Structure, Different Surface
Give three problems from different contexts that share one underlying relationship. Students identify the invariant before solving.
Training Drill 2 | Same Surface, Different Structure
Give near-twin questions that look almost identical but require different methods because one condition changes. Students identify the discriminating condition before execution.
Training Drill 3 | Strip the Story
Remove names and context from a word problem until only quantities, relationships, constraints and target remain. Students state the structural skeleton, then restore the story and solve.
Training Drill 4 | Change the Representation
Convert the same problem among prose, table, graph, diagram and equation where meaningful. Students explain what information remains invariant across forms.
Training Drill 5 | Reverse the Direction
Turn cause-to-effect questions into effect-to-cause questions, output-to-parameter questions, or evidence-to-inference questions into claim-to-evidence questions. Transfer strengthens when relationships work both ways.
Training Drill 6 | Command Swap
Keep the same content and switch describe to explain, explain to evaluate, find to show or compare to judge. Students identify what content transfers and what answer operation must change.
Training Drill 7 | Boundary Transfer
After solving an ordinary case, test zero, equal, empty, maximum or threshold cases where appropriate. Students state which hidden assumptions the boundary exposes.
Training Drill 8 | Analogy Map
Create two columns: old problem and new problem. Map target, givens, relationships, constraints and answer form. Leave surface details unmapped unless they actually matter.
Training Drill 9 | Analogy Break
Give an analogy that works in several respects and fails in one method-critical respect. Students identify where transfer must stop.
Training Drill 10 | Method Permission
For each known method, students list the minimum conditions required for use. Transfer practice then varies those conditions deliberately.
Training Drill 11 | Structure Naming
After solving, students name the problem family without using the chapter title. Examples: conserved total with redistribution, multiplicative comparison, causal chain with confound, evidence-to-inference, state update with invariant.
Training Drill 12 | Transfer Explanation
Students complete: “This method transfers because ___ remains the same. I must change ___ because ___ changed.” The sentence forces separation of invariant and adaptation.
Training Drill 13 | Mixed Surface Set
Mix questions from several chapters and representations. Students must identify structure before method. No chapter labels are shown.
Training Drill 14 | False Friend Set
Use questions containing familiar keywords that point toward the wrong method unless conditions are checked. The learner practises resisting keyword transfer.
Training Drill 15 | Cross-Subject Relation
Compare one Mathematics proof, one Science explanation and one English inference as premise/evidence → transformation → conclusion. Students identify the shared reasoning architecture without pretending the subject-specific standards are identical.
Training Drill 16 | Delayed Transfer
Teach a method, wait, then present it in a changed surface without announcing the connection. Delayed transfer tests whether the structure can be retrieved rather than merely kept active in working memory.
Training Drill 17 | Timed Transfer
After untimed transfer is stable, add realistic time. Students must perform the structural check quickly enough to be useful in examination conditions.
Training Drill 18 | Wrong Transfer Diagnosis
Give a solution where an old method was transferred incorrectly. Students identify which condition failed and what parts of the old method, if any, still survive.
Training Drill 19 | Minimal Adaptation
Students take a known method and modify only the steps that the new condition requires. This prevents both total restart and blind copying.
Training Drill 20 | Transfer Log
Record unfamiliar questions that became solvable after structural mapping. Note old method, invariant, changed feature, adaptation and verification. Over time the learner builds a library of transferable structures rather than a library of isolated questions.
A One-Week Transfer Programme
Day 1: distinguish surface from structure. Day 2: same structure/different surface. Day 3: same surface/different structure. Day 4: representation switching. Day 5: direction and command changes. Day 6: mixed unfamiliar questions. Day 7: timed transfer and error review.
A Four-Week Integration Programme
Week 1: identify invariants and method conditions. Week 2: map across representations and contexts. Week 3: train near twins, boundaries and false analogies. Week 4: mixed timed papers where transfer opportunities are unannounced.
What to Measure
- same-surface accuracy;
- changed-surface accuracy;
- representation-transfer accuracy;
- method-selection accuracy without chapter labels;
- near-twin discrimination;
- false-transfer errors;
- time to identify the invariant;
- ability to state method conditions;
- successful adaptation rather than full restart;
- performance on delayed unfamiliar questions.
The target is not to make every question feel familiar.
It is to make unfamiliarity less important than structure.
Using AI to Train Transfer Without Outsourcing the Mapping
- “Give me three questions with the same deep structure but very different surface contexts. Do not name the common method.”
- “Give me two near-twin questions where one condition changes the correct method.”
- “After I solve this, change only the representation and ask me to map the old roles to the new form.”
- “Reverse the direction of this problem without changing the underlying relationship.”
- “Give me an analogy that works in several ways but fails in one method-critical way.”
- “Ask me which parts of my old solution transfer unchanged and which must be adapted.”
- “Do not tell me the chapter. Ask me to classify the structural problem family from givens, target and constraints.”
AI is especially useful for generating controlled variation. The learner should still perform the abstraction, mapping and condition check.
The AI Surface Trap
AI can generate many paraphrases that look different while preserving almost everything else.
That can create shallow “variation” practice.
Ask for controlled structural variation, not cosmetic rewording only.
Change representation, direction, boundary, command, assumption or evidence status.
Real transfer training varies what matters.
The Examination-Day Micro-Routine
What is the job? What relationship is invariant? What changed? Do the old method’s conditions still hold? Map the roles. Adapt only what changed. Go.
Frequently Asked | What If the Question Looks Completely New?
Strip the context. Write target, givens, relationships and constraints. Ask what known structure could connect them. A new story can conceal an old relationship. If no structural match appears, generate a route from first principles rather than forcing familiarity.
Frequently Asked | What If the Question Looks Exactly Like Practice?
Use familiarity as a useful hypothesis, not automatic permission. Check whether the structure, conditions and target actually match. Near twins are common sources of transfer error because one small changed condition can invalidate the old route.
Frequently Asked | Is Transfer the Same as Analogy?
Analogy is one route into transfer. It maps relational structure between cases. Transfer is broader: it includes reusing methods, representations, criteria, checking routines and causal structures when the relevant conditions hold.
Frequently Asked | Is Transfer Just Practice?
Practice is necessary and can remain too narrow. Transfer requires variation that forces the learner to recognise what is invariant and what changes. Repeating one surface pattern can improve speed without creating flexible application.
Frequently Asked | Should I Memorise Templates?
Templates can reduce load when they encode real structure. Learn the conditions and roles behind them. A template that cannot be adapted when the question changes is not transferable knowledge; it is a brittle script.
Frequently Asked | Does This Apply to Primary Students?
Yes. Use simple prompts: “What is the same?” “What changed?” “Does our old method still work?” “Which part must we change?” Primary transfer can begin with number bonds, bar models, measurement, grammar, comprehension clues and familiar Science mechanisms in changed contexts.
Frequently Asked | Does This Apply at University?
Yes. Advanced study depends heavily on transfer across novel proofs, cases, datasets, models, codebases, research questions and professional contexts. The conditions become more specialised, but the invariant/adaptation distinction becomes even more important.
Canonical Owner Boundaries
This article owns examination-method transfer across changed question surfaces: identifying invariant structure, mapping new problem roles onto a known method, checking conditions of use and adapting only the parts genuinely changed by the new problem.
- How Intelligence Works | Transfer and Recomposition owns the broad cognitive architecture of transfer across changed problems.
- How Intelligence Works | Abstraction owns the general operation of stripping away surface detail to preserve structure.
- How Intelligence Works | Analogy owns structural analogy across domains.
- Know When a Familiar Pattern Is Misleading owns false familiarity and near-twin traps.
- Know When to Slow Down Before You Commit owns tempo control at high-risk transfer decisions.
The next edge, Think Across Mathematics, Science and English, will own cross-domain reasoning operations that appear across subject boundaries. This page stays narrower: carrying one method or reasoning structure across a changed examination problem without mistaking surface resemblance for structural equivalence.
Evidence and Limits
Transfer is difficult precisely because knowledge can become tied to the contexts in which it was learned. Surface variation, mixed practice, comparison, explanation of conditions and representation switching can all help learners recognise deeper structure, but no single exercise guarantees broad transfer. Training should progress from supported near transfer toward less cued, more varied and delayed transfer.
There is also a real limit to transfer. Some methods are domain-specific because their assumptions belong to the domain. A causal standard in Science is not identical to literary interpretation. Mathematical proof standards are not the same as source evaluation. Cross-domain analogy can illuminate structure and should not erase disciplinary rules.
The strongest transferable knowledge therefore includes both the reusable structure and the conditions that limit reuse.
The World Return
Outside examinations, transfer is one of the central ways expertise becomes useful.
An engineer carries a conservation principle into a new machine while rechecking boundary conditions.
A doctor recognises a familiar physiological pattern in a patient whose presentation looks different while refusing to ignore patient-specific evidence.
A programmer reuses an algorithmic idea in a different codebase while adapting interfaces and constraints.
A lawyer recognises a reasoning structure in a new fact pattern while checking whether the governing rule and jurisdiction still apply.
A teacher carries a diagnostic principle from one subject to another while respecting subject-specific knowledge.
Expertise is not having a separate memorised answer for every possible situation.
It is having structures that can travel—and knowing the border where they must stop.
Transfer is not copying the past into the present. It is carrying forward what remains true and rebuilding what changed.
The Return to the Table
Jo gives the group six unfamiliar questions.
No chapter headings.
No familiar worksheet formatting.
No clue which examples they resemble.
Ben sees a familiar pattern and checks its conditions before committing.
Aisha reconstructs the relationship from target and givens when no keyword appears.
Ryan tests the closest structural match once and stops generating alternatives.
Mira maps old roles to new roles on the page.
Clara strips away the story until the invariant becomes visible.
Ethan sees a powerful analogy and asks where the analogy stops.
The questions remain unfamiliar.
The students are no longer helpless in the face of unfamiliarity.
The surface can change completely. If the structure survives, the method can travel. If the structure changes, intelligence travels by adapting—not by copying.
Advanced Transfer Atlas | 45 Changed-Surface Cases
The cases below deliberately vary what students usually mistake for the topic. Each case asks the same four questions: what changed on the surface, what stayed invariant, what condition gives the old method permission, and what adaptation is genuinely necessary?
Atlas 1 | Percentage Change Without the Word Percentage
A question asks how much larger a new quantity is relative to an old quantity without using the phrase “percentage increase.” The old method still transfers because the invariant job is multiplicative comparison against a reference base. The surface language changed; the reference relationship did not. The student should map old value to base, new minus old to change, and preserve the direction of comparison.
Atlas 2 | Same Formula, Different Unknown
Practice always supplied radius and asked for area. The exam supplies area and asks for radius. The formula transfers, but the direction of solving reverses. The invariant is the area-radius relationship; the adaptation is algebraic inversion and domain interpretation. Students who learned only forward substitution may mistake this for a new topic.
Atlas 3 | Same Ratio, Units Change
A familiar speed relationship appears with minutes and kilometres rather than hours and metres. The ratio structure transfers; the units require adaptation. The method has permission only after quantities are placed in compatible units. Blind reuse of numeric patterns without unit mapping produces false transfer.
Atlas 4 | Same Geometry, Rotated Diagram
The triangle is turned ninety degrees and the student no longer recognises it. Orientation is surface. Angle relations and side correspondence are structural. Redraw, relabel or mentally rotate if helpful, but do not relearn the theorem. Transfer means recognising the geometry independent of page orientation.
Atlas 5 | Same Geometry, One Condition Removed
The new diagram looks almost identical to practice but no longer provides parallel lines. The surface invites transfer while the structural permission disappears. Preserve all general geometry knowledge, but reject the corresponding-angle route. This is the canonical warning that similarity of appearance is weaker than similarity of conditions.
Atlas 6 | Linear Pattern Hidden in a Table
Students practised straight-line equations from graphs. The exam gives a table. The transferable invariant is constant rate of change. Compute differences, identify the relationship and reconstruct the line. Representation changed; linear structure survived.
Atlas 7 | Quadratic Pattern Hidden in a Context
A rectangular design problem creates a product of two changing dimensions. No quadratic keyword appears. The structure emerges only after modelling. Transfer requires building the representation first, then recognising the algebraic family. The method cannot transfer before the student has exposed the structure.
Atlas 8 | Same Probability, Replacement Changes
The story and objects are nearly identical to practice, but sampling changes from with replacement to without replacement. The tree-diagram method can still transfer, but branch probabilities must adapt because state changes. The invariant is conditional branching; the surface similarity must not erase the changed sample space.
Atlas 9 | Same Probability, Order Becomes Irrelevant
A student learned to multiply path probabilities in ordered sequences. The new question asks for a combination where order does not matter. The probability principles remain, but counting structure changes. Transfer requires noticing whether permutations are distinct outcomes under the new target.
Atlas 10 | Same Algebra, Context Imposes Domain
An equation has two algebraic roots, but only one can represent a time, length or count. The solving method transfers unchanged; answer validation adapts to context. Domain restrictions are part of transfer because real-world interpretation can change which mathematically valid candidate survives.
Atlas 11 | Same Conservation Principle, New Apparatus
A Science question replaces the familiar laboratory apparatus with an unfamiliar device. The invariant is conservation of the relevant quantity; the surface hardware changes. Students should map inputs, transfers and outputs before deciding what equation or explanation transfers.
Atlas 12 | Same Mechanism, Different Measured Outcome
The underlying process is unchanged, but the experiment now measures gas volume instead of mass loss. The mechanism transfers; the observable consequence must be remapped. Strong transfer keeps causal structure while changing the indicator used to detect it.
Atlas 13 | Same Trend, Different Evidence Strength
Two datasets show similar upward relationships. One comes from a controlled experiment; the other from observational records. Trend-reading transfers, but causal inference does not. Evidence status is part of the method’s condition set.
Atlas 14 | Same Experimental Flaw, Different Variable
Students learned that uncontrolled temperature could confound an experiment. A new task has uncontrolled light intensity. The variable name changes, but the transferable structure is “a second variable differs systematically between conditions and could explain the outcome.” The diagnostic logic transfers across variables.
Atlas 15 | Same Model, Boundary Condition Changes
A model works for ordinary temperatures and the new problem includes an extreme range where assumptions fail. Transfer must stop at the model boundary. Strong students learn not only a model but the conditions under which the model is licensed.
Atlas 16 | English Inference Moves From Fiction to Non-Fiction
The student learned inference through character actions in fiction. A speech now requires inference about speaker attitude. The genre changes; evidence-to-inference remains. Transfer succeeds when the learner treats wording, emphasis and selection as evidence rather than searching for narrative clues that no longer exist.
Atlas 17 | Same Quotation Method, Different Effect
The student knows to select a precise quotation and explain it. That method transfers. The memorised effect does not. The new wording must generate a new interpretation. Structural transfer preserves the analysis operation while rebuilding content.
Atlas 18 | Same Evidence, Command Changes
A passage supports the same idea, but the command changes from “what does this suggest?” to “how does the writer create this impression?” The evidence can transfer. The answer operation must add method and effect. Transfer is partial rather than all-or-nothing.
Atlas 19 | Same Topic, Summary Instead of Analysis
The learner recognises the topic and begins analysing language. The new task asks for summary. Knowledge of the passage transfers; analytical operations must be suppressed. Strong transfer includes knowing what not to carry across.
Atlas 20 | Comparison Dimension Changes
Practice compared tone. The exam asks to compare purpose. The general comparison method transfers: align both texts on a shared dimension, use evidence and make the relation explicit. The specific dimension changes and therefore the evidence selected must change too.
Atlas 21 | Historical Cause Framework Moves to a New Topic
The factual causes differ completely. The temporal architecture—background conditions, trigger, mechanisms and consequences—can transfer. This allows students to organise unfamiliar content without pretending the actual history is the same.
Atlas 22 | “Most Important” Moves Across Topics
The factors change, but the evaluative operation persists: select a criterion, compare the strongest contenders and justify the ranking. The transferable object is not a memorised factor list. It is the judgement architecture.
Atlas 23 | Source Reliability Moves to Source Usefulness
Some source-analysis habits transfer: inspect content, provenance, purpose and claim. The criterion changes. Reliability asks whether the source can be trusted for a specific claim; usefulness asks what it contributes to the inquiry. Transfer must preserve the evidence-analysis structure while adapting the judgement target.
Atlas 24 | Change and Continuity Moves Across Timeframes
The same comparison method can apply over ten years or one hundred years, but the scale of evidence and meaning of continuity change. Transfer requires rechecking timeframe and criterion before reusing paragraph structure.
Atlas 25 | Algorithm Moves From Numbers to Records
A maximum-finding loop transfers from numeric lists to records containing scores. The invariant is “best candidate seen so far.” The adaptation is the comparison key. Syntax and data type change; state logic survives.
Atlas 26 | Algorithm Moves From Fixed to Empty-Possible Input
The old method assumed at least one item. The new specification permits empty input. Core iteration logic transfers, but initialisation and guard conditions must change. Transfer without boundary adaptation creates a hidden failure.
Atlas 27 | Search Algorithm, Different Ordering Assumption
The learner knows binary search and sees a search problem. The new data are unsorted. The surface goal matches; the structural precondition does not. The method must not transfer until ordering is established or a different search route is selected.
Atlas 28 | Same Test Strategy, Different Failure Mode
Boundary testing transferred from one program to another, but the new program’s main risk is duplicate input rather than range endpoints. The general idea “choose a test that attacks plausible failure” transfers; the actual test case must be rebuilt from the new vulnerability.
Atlas 29 | Same Statistical Tool, Different Assumptions
A familiar statistical method appears attractive because the data look similar. Transfer requires checking its assumptions, not only its formula. If distribution, independence, scale or sampling conditions change, the old method may need adaptation or rejection.
Atlas 30 | Same Proof Pattern, Different Domain
A contradiction proof pattern can transfer, but the specific assumptions and allowed operations depend on the new domain. Students should carry the proof architecture—assume opposite, derive impossibility, conclude—but rebuild the domain-specific bridge.
Atlas 31 | Same Error-Checking Routine, Different Subject
In Mathematics the learner checks a candidate against the original equation. In English the analogous operation is checking an inference against decisive textual evidence. The exact standards differ, but the meta-structure “return candidate to source constraints” can transfer across subjects.
Atlas 32 | Same Elimination Logic Across Domains
Multiple-choice elimination removes options violating explicit constraints. In Mathematics a root violates domain; in Science a hypothesis contradicts data; in English a reading contradicts a key phrase. The transferable structure is constraint-based pruning, not a subject-specific trick.
Atlas 33 | Same Confidence Calibration, Different Evidence
Confidence should track evidence in every subject, but evidence has different forms. In Mathematics it may be derivation and substitution; in Science controlled data; in English textual support; in Humanities corroboration and criteria. The calibration principle transfers while the evidence standard remains domain-specific.
Atlas 34 | Same Problem-Solving Skeleton, Different Representation
The learner uses target → givens → constraints → relationship → method in Mathematics. The same skeleton can help organise an unfamiliar Science or Computing task, but “relationship” and “method” must be interpreted in domain terms. Transfer is useful when abstraction remains humble enough to respect disciplinary meaning.
Atlas 35 | Same Concept, Different Scale
A pattern learned in a small example is applied to a much larger system. The method transfers only if scale does not change the governing assumptions. Students should ask whether new interactions, approximation errors or constraints appear at the larger scale.
Atlas 36 | Same Surface Keyword, Different Operation
The word “rate” appears in two questions. One asks for a ratio; another asks for rate of change from a graph. Keyword matching alone is unsafe. Transfer requires inspecting the target and relationship rather than retrieving a method from one word.
Atlas 37 | Different Surface Keyword, Same Operation
One question says “per,” another “for every,” another gives a gradient, and another describes constant change verbally. The common structure may be a rate. Strong transfer recognises synonyms and representational variants without relying on exact phrasing.
Atlas 38 | Same Worked Example, Numbers Rearranged
The problem contains the same numerical values as practice but assigns them to different roles. Surface memory becomes dangerous. Map quantities by meaning, not position or number identity. Transfer should follow roles rather than copied slots.
Atlas 39 | Same Structure, Distracting Extra Information
The exam adds irrelevant details around a familiar relationship. The invariant method still transfers. The adaptation is information selection: distinguish givens that constrain the target from decoration or context that does not enter the model.
Atlas 40 | Same Structure, Missing One Familiar Cue
The usual diagram, keyword or first step is absent. Instead of concluding the method is new, reconstruct from the relationship. Transfer competence is strongest when the learner can regenerate the cue from structure rather than requiring the cue to be supplied.
Atlas 41 | Same Method, Different Answer Form
The reasoning produces a ratio, but the new task requests a percentage. The internal method transfers; the final conversion changes. Transfer includes the output interface. A correct route can still fail if the student copies the old answer form.
Atlas 42 | Same Evidence, Different Audience
A writing task asks the student to reuse knowledge for a different audience and purpose. Content can transfer, but selection, explanation depth, tone and organisation may need adaptation. Transfer is not identical repetition across communication contexts.
Atlas 43 | Same Diagnostic Routine, New Error Family
The learner knows to find the first weak link after mistakes. In a new subject, the diagnostic categories differ. The meta-routine transfers—locate first divergence, classify, repair—but the possible failure modes must be learned within the domain.
Atlas 44 | Same Planning Routine, New Constraint
A planning framework developed for untimed essays is used under a strict timed response. The architecture can transfer, but depth must compress. Transfer includes resource adaptation; a method that ignores the new time constraint is not fully transferred.
Atlas 45 | The Method Should Not Transfer at All
The strongest transfer skill sometimes produces the answer “none of my familiar methods has permission yet.” This is not failure. It is disciplined refusal to force an analogy. Return to target, givens and constraints, build a new representation and generate a route from the live problem.
The Atlas Pattern
Across all forty-five cases, successful transfer carries forward a role structure, not a visual memory. It asks which relationship remains, which conditions license reuse, which new features matter and which old details should be discarded. Transfer therefore has two equal responsibilities: recognise when knowledge travels and recognise where it must stop.
Carry the invariant, not the costume.
Deep Transfer Lab | 35 Diagnostic Repairs
Transfer errors need precise diagnosis. A student who cannot see the old structure inside a new surface needs a different intervention from a student who sees the structure but ignores a changed condition. The cases below isolate the first weak link so practice can target the actual failure instead of assigning more of the same worksheet.
Diagnostic 1 | Surface Lock
The student solves a method fluently in the exact format practised and fails when the story changes. Strip both old and new problems to target, givens, relationships and constraints. Ask the learner to identify what remains the same before solving. Do not add more same-surface repetitions; add controlled variation.
Diagnostic 2 | Keyword Dependence
The learner can solve “percentage increase” questions but fails when the phrase is absent. Train paraphrase families and relationship-first classification. Hide topic labels and keywords. Ask what the quantity is relative to, not what chapter the question belongs to.
Diagnostic 3 | Template Copying
The student reproduces a familiar sequence even when one step has no purpose in the new question. Ask them to explain the function of each template step. Then present a near twin where one step must be removed or replaced. Transfer becomes functional rather than visual.
Diagnostic 4 | Condition Loss
The old method is remembered but its permission conditions are not. Build method cards with two sides: what the method does and when it is allowed. Practice should vary the permission condition deliberately. The learner must sometimes reject a familiar method.
Diagnostic 5 | Over-Abstraction
The learner sees a deep pattern and ignores a changed detail that actually matters. Train “invariant versus discriminating change.” After naming the common structure, require the student to identify the one new feature most capable of invalidating the old route. Abstraction should remove noise, not erase conditions.
Diagnostic 6 | Under-Abstraction
Every new context feels like a new topic. Ask the learner to rename quantities generically—input, output, rate, total, constraint, evidence, claim—then compare with a known example. Generic role language can reveal structural identity hidden by story detail.
Diagnostic 7 | Representation Lock
The learner knows a relationship only as a formula or only as a diagram. Use representation-switch drills. Ask for graph, table, verbal rule, equation or diagram versions of the same structure. The goal is not artistic variety; it is structural invariance across formats.
Diagnostic 8 | Direction Lock
The learner can compute output from input but cannot solve for input from output. Reverse familiar questions. Ask for causes from effects, parameters from results, evidence from claims and conditions from outcomes. A relationship owned in one direction is only partially transferable.
Diagnostic 9 | Command Lock
The learner recognises topic content but writes the same response structure whether asked to describe, explain, evaluate or compare. Keep the stimulus constant and change only the command. Students should see that content transfer can coexist with operation change.
Diagnostic 10 | Boundary Blindness
The method works on ordinary examples and fails at zero, empty input, equality or thresholds. Add boundary variants after each routine family. Ask what hidden assumption the boundary exposes. Transfer becomes robust when the learner knows the edge of method validity.
Diagnostic 11 | Analogy Overreach
The learner maps too many features from the old problem. Use analogy maps with two columns: necessary correspondences and incidental similarities. Ask which mapped features the method actually depends on. Transfer should preserve relations, not decorative resemblance.
Diagnostic 12 | Analogy Underuse
The learner refuses to use an old method because the new context feels unrelated. Ask for one relational sentence: “Both problems involve ___ changing with ___ under ___ constraint.” If that sentence can be stated accurately, test the mapping instead of abandoning the connection.
Diagnostic 13 | Transfer Only After a Hint
The teacher says “this is like the ratio problem” and the learner succeeds. The support is carrying classification. Fade the hint: first name the family, then ask only “what stays the same?”, then provide no cue. Transfer readiness grows as classification becomes independent.
Diagnostic 14 | Transfer Only Immediately After Practice
The learner succeeds while the method is still active in working memory and fails later. Add delayed changed-surface questions. Space the transfer attempt so retrieval of structure is required. Immediate success is weaker evidence of durable transfer.
Diagnostic 15 | Transfer Fails in Mixed Sets
Single-topic worksheets are strong; mixed papers collapse. The missing layer is method selection. Interleave neighbouring problem families and remove chapter headings. Require structural classification before execution. Mixed practice should be introduced after each method is sufficiently understood, not before.
Diagnostic 16 | Transfer Fails Under Time
Untimed structural mapping is accurate but too slow. Identify which transfer operation consumes time: stripping surface, retrieving old structure, checking conditions or mapping roles. Compress that operation through focused practice. Do not remove the condition check merely to gain speed.
Diagnostic 17 | False Transfer Under Time
Time pressure strengthens first-pattern matching. Use timed near-twin sets where one changed condition matters. Train a three-second rule: same job, same relationship, same conditions? If yes, go. If not, adapt.
Diagnostic 18 | Transfer Fails After Surface Decoration
Extra names, diagrams, story details or numbers distract the learner from the core. Teach information triage: target, givens, constraints, relationship. Ask which details would change the solution if removed. Irrelevant decoration should gradually lose control over method selection.
Diagnostic 19 | Transfer Fails When Familiar Cue Is Missing
The learner relies on a phrase, diagram or first step supplied by the worksheet. Remove the cue and ask the learner to regenerate it. If the diagram is necessary, draw it. If a variable definition is necessary, create it. Transfer improves when useful cues become outputs of reasoning rather than prerequisites.
Diagnostic 20 | Method Transfers, Answer Form Does Not
The reasoning is correct and the student submits the old answer form. Train final-role remapping. Before execution, write the new target object: value, percentage, comparison, judgement, vector, probability or explanation. Method transfer must terminate in the new interface.
Diagnostic 21 | Content Transfers, Evidence Standard Does Not
A general reasoning principle crosses subjects and the learner assumes the evidence requirements also cross unchanged. Compare domain standards explicitly. Mathematical proof, scientific causal evidence and literary inference each demand different warrants. Cross-domain transfer should preserve meta-structure while respecting local epistemology.
Diagnostic 22 | Same Subject, Wrong Level of Generality
The learner transfers a narrow shortcut as if it were the general method, or uses a very general method when a direct specialised route exists. Ask what problem family the method actually covers. Build nested knowledge: principle, common method, special case, shortcut.
Diagnostic 23 | Student Cannot State the Invariant
The learner can solve the old and new question but cannot explain why the method transfers. Ask for one sentence: “This works in both because ___.” If the sentence is only “they are similar,” press for the relationship or condition. Naming the invariant makes future transfer more deliberate.
Diagnostic 24 | Student States an Invariant That Is Too Vague
“Both use maths,” “both are about rates,” or “both need evidence” may be true and unhelpful. Increase precision: constant multiplicative ratio, change per unit input, evidence supports inferred attitude, source purpose changes reliability for claim X. Useful abstraction is neither concrete surface nor empty generality.
Diagnostic 25 | Student States an Invariant That Is Too Narrow
The learner describes the exact numbers or nouns rather than the relation. Replace specifics with roles. “Three apples for two oranges” becomes “two corresponding quantities in a fixed ratio.” The abstraction should survive a different context.
Diagnostic 26 | Student Adapts Too Much
A changed surface makes the learner rebuild the whole method from scratch. Ask which old steps remain valid under the new conditions. Preserve them. Transfer should reduce cognitive cost by carrying stable structure forward.
Diagnostic 27 | Student Adapts Too Little
The old sequence is copied exactly despite a changed boundary, command or variable role. Highlight the discriminating change and require one sentence explaining which step it affects. Adaptation should be proportional to structural change.
Diagnostic 28 | Student Transfers the Example, Not the Check
The old solution includes a useful verification, but the learner copies only the solving steps. Teach check transfer explicitly. Ask what failure mode the old check attacked and whether the same failure remains plausible. If yes, transfer the check or design an equivalent one.
Diagnostic 29 | Student Transfers the Check When Risk Changed
The learner always uses the same verification even when the new problem’s vulnerability differs. Transfer the principle “attack the dominant failure,” not the literal check. New risk may require a new diagnostic test.
Diagnostic 30 | Student Transfers Across Subjects Too Aggressively
Meta-routines such as compare alternatives or test constraints are useful across subjects, but the learner begins treating all evidence as if it were the same. Reinforce domain boundaries. Cross-subject transfer should happen at the reasoning-operation level while local criteria remain subject-specific.
Diagnostic 31 | Student Never Transfers Across Subjects
The learner sees Mathematics checking, Science evaluation and English evidence analysis as unrelated. Compare their shared skeletons without erasing differences. For example: candidate → source constraints → judgement appears in each domain. This can build general self-regulation while retaining subject standards.
Diagnostic 32 | Student Uses Transfer to Avoid Learning New Content
A general framework is applied where missing domain knowledge is the real blocker. Clarify that transfer cannot manufacture facts, vocabulary, theorem statements or scientific mechanisms that were never learned. Transfer organises and reuses knowledge; it does not replace acquisition.
Diagnostic 33 | Student Learns Content Without Building Transfer Hooks
Knowledge remains isolated in chapter-specific memory. After teaching, ask: what other problems use this relationship? What condition would break it? What representation could show the same idea? What neighbouring method is easily confused with it? These questions attach structural hooks for future retrieval.
Diagnostic 34 | Transfer Success Is Not Recorded
The learner solves an unfamiliar question and moves on without extracting the lesson. Record the transfer: old structure, changed surface, invariant, adaptation, verification. The log helps turn isolated success into a reusable transfer library.
Diagnostic 35 | Transfer Failure Is Labelled “Careless”
A false analogy, missed condition or representation lock is called carelessness. Replace the label with a precise transfer failure category. The corrective drill should target the mechanism: near twins, condition cards, representation changes, command swaps or boundary cases.
The Transfer Design Protocol
When creating practice, vary one structural dimension at a time before combining several. A useful sequence is: numbers → wording → context → representation → direction → command → boundary → assumption → mixed selection. Controlled variation lets teacher and learner see exactly where transfer breaks.
The Contrast-Pair Protocol
Transfer improves when learners compare two cases deliberately. Use one pair where surface changes but structure stays, and another where surface stays but structure changes. Ask what cue should drive method choice. Contrast sharpens decision boundaries better than repetition alone.
The Transfer Card
For high-value methods, create a compact card:
Job → invariant → permission conditions → common surface changes → failure boundary → best check.
This is more useful than a card containing only a formula or template because it includes the knowledge required to decide when reuse is legitimate.
The Primary-School Transfer Ladder
Keep transfer concrete: same idea, new picture; same method, new numbers; same clue, new story. Ask “What is the same?” and “What changed?” before solving. Use bar models, number bonds, measurement, grammar patterns, comprehension evidence and Primary Science mechanisms across varied contexts.
The Secondary-School Transfer Ladder
Secondary learners should add method conditions, representation changes, reversed directions, near twins, mixed-topic selection and answer-command changes. They should increasingly classify by structure rather than chapter label.
The JC, IB and University Transfer Ladder
Advanced transfer includes model assumptions, proof forms, uncertainty, source standards, statistical conditions, code architecture and novel cases. The cost of over-transfer rises because methods become more powerful and more conditional. Expertise means carrying abstractions together with their limits.
Teacher Protocol | Ask What Survives the Change
When a learner says “this is different,” ask which relationship remains. When they say “this is the same,” ask which condition could break the analogy. These two questions train both sides of transfer: seeing through surface change and respecting structural change.
Tutor Protocol | Fade Topic Labels
Do not let every worksheet announce the method family forever. Once foundations are secure, mix neighbouring families and remove labels. Ask the student to justify classification before execution. Tutor support should eventually shift from “this is a ratio question” to “what relationship do you see?”
Parent Protocol | Ask for One Different Example
After homework, parents can ask: “Can you make up a different-looking question that uses the same idea?” or “What could change and make this method stop working?” The parent need not know the full solution. The questions encourage structural ownership.
Performance Dashboard
Track same-surface accuracy, changed-surface accuracy, near-twin discrimination, representation transfer, delayed transfer, mixed-set method selection, false-transfer errors, condition-check accuracy, adaptation quality and time to recognise invariant structure. Improvement should appear as a narrowing gap between familiar and unfamiliar performance.
Master Transfer Routine
Strip the surface. Name the job. Find the invariant. Check permission. Map roles. Adapt the changed parts. Verify against the new problem.
Final Principle
Unfamiliar questions are difficult when knowledge is stored as a collection of surfaces. They become more manageable when knowledge is stored as relationships, roles, conditions and boundaries. But abstraction alone is not enough. The student must also notice when one changed detail destroys the old method’s permission.
The best transfer does two things at once: it sees sameness beneath difference, and difference inside sameness.