Maya used to get this right.
That is what makes the mistake interesting.
Three weeks ago, percentage change was stable.
Original value.
Difference.
Divide by the reference quantity.
She could explain why.
She could do it cold.
She could do it in mixed questions.
Then school introduced percentage error.
Now Maya keeps dividing by the wrong quantity in both topics.
Her mother says:
“She forgot percentages again.”
Maybe.
But the timing is suspicious.
The old skill did not weaken quietly during a long gap.
It became unreliable immediately after a similar new rule arrived.
The tutor writes two questions.
One asks for percentage change.
One asks for percentage error.
No calculations yet.
Only this:
What is the reference quantity in each problem, and why?
Maya pauses.
Then says:
“I think I mixed the two rules.”
Exactly.
The old learning may still be present.
The new learning may also be present.
The problem may be that both are trying to answer the same cue.
That is interference.
And tuition needs a way to test for it.
When an old capability suddenly becomes unreliable after similar new learning appears, ask whether the learner has forgotten the old rule—or whether two rules are now competing for the same situation.
That is the Interference Test.
Quick Read: The Interference Test in One Sentence
If a once-stable skill fails soon after similar learning is introduced, compare the old and new rules directly, isolate their distinguishing cues, and test whether the learner can select between them before reteaching either one from zero.
This article owns a narrow learning job.
The Maintenance Window asks when a stable skill should return.
The Reactivation Cost asks how hard it is to restore a faded skill.
The Interference Test asks something different:
Did the old capability decay, or did a newer, similar capability begin competing with it?
1. Forgetting and Interference Can Look Identical
From the outside, both can produce the same event.
The student gets the old question wrong.
But the mechanisms are different.
Decay story:
The old knowledge has become difficult to retrieve because it has not been used enough.
Interference story:
The old knowledge remains available, but another similar response is being activated by the same or overlapping cues.
These stories suggest different repairs.
Decay may need retrieval and reactivation.
Interference may need discrimination.
Reteaching both topics separately can leave the competition untouched.
2. Similarity Is the First Clue
Interference becomes more plausible when the old and new learning resemble each other.
Two formulae with similar symbols.
Two grammar rules with overlapping sentence forms.
Two Science concepts using related vocabulary.
Two differentiation rules that can appear in the same expression.
Two percentage relationships using the same numbers but different reference quantities.
Similarity is useful for learning because it allows transfer.
Similarity can also create competition because the learner must notice the feature that tells the cases apart.
The educational problem is often not:
“Does the student know Rule A?”
or:
“Does the student know Rule B?”
It is:
“Can the student tell when A rather than B is the correct response?”
3. Timing Is the Second Clue
A skill has been stable for months.
Then a related topic is taught.
Within days, the old errors appear.
That temporal pattern should raise the interference hypothesis.
It does not prove interference.
The new topic may coincide with fatigue, examination stress or a long gap in old practice.
But the timing tells the tutor where to look.
This is where The Flight Recorder matters.
Without history, the tutor may forget that the old skill was recently strong.
4. Error Substitution Is the Third Clue
Ordinary forgetting often produces hesitation, blankness or incomplete retrieval.
Interference often produces something more revealing:
the learner confidently applies the wrong but related rule.
Maya does not say:
“I have no idea.”
She uses the percentage-error denominator inside a percentage-change question.
Hana does not forget every writing structure.
She applies the situational-writing tone to a different text type.
Jia Jun does not stop reasoning scientifically.
He substitutes condensation language into an evaporation explanation because both involve water and state change.
The presence of a coherent competitor is evidence worth investigating.
5. The Interference Test Begins With a Pair
Do not start with twenty questions.
Start with two.
One old case.
One new case.
Keep them as similar as possible except for the feature that determines the correct rule.
Then ask:
What tells you these two problems are different?
If the learner can state the discriminating feature, both rules may be intact but insufficiently self-triggered.
If the learner cannot explain the difference, the conceptual boundary may never have been encoded clearly.
6. Separate Rule Knowledge From Rule Selection
Give Rule A explicitly.
Can the student execute it?
Give Rule B explicitly.
Can the student execute it?
If yes to both, the problem may not be knowledge of either rule.
Now remove the labels and mix them.
If performance collapses only there, selection and discrimination become the primary learning job.
This is a classic example of why a Diagnostic Probe can prevent unnecessary reteaching.
7. The Contrast Question
A high-value interference question is:
Why is Rule A correct here but Rule B wrong?
This forces the learner to encode the boundary, not merely rehearse the response.
For percentage work:
Which quantity defines the comparison?
For grammar:
Which sentence feature changes the rule?
For Science:
Which direction of process distinguishes evaporation from condensation?
For Additional Mathematics:
Which operation is outermost and therefore determines whether Product Rule, Chain Rule or both are needed?
Contrast is not extra decoration.
Contrast is the learning object.
8. Proactive and Retroactive Interference Are Useful Labels
Memory research often distinguishes two directions of interference.
Proactive interference:
older learning disrupts newer learning.
Retroactive interference:
newer learning disrupts access to older learning.
Real school learning can be messier than these clean labels.
But the distinction is useful because competition can run in both directions.
An old formula can intrude into a new topic.
A newly learned procedure can then begin contaminating the old one.
The tutor should inspect both.
9. The Learner May Know Both Rules and Still Fail
This is one of the most important points.
A student can correctly explain Rule A in isolation.
Correctly explain Rule B in isolation.
And still choose the wrong one when the two are mixed.
This is why blocked practice can create misleading confidence.
When every question belongs to one rule family, the learner does not need to discriminate.
The worksheet has already performed the selection.
Interleaving different problem types can increase the need to identify which strategy fits each problem. Classroom research in Mathematics has found benefits from this kind of interleaved practice, including improved strategy discrimination compared with blocked practice.
10. Interleaving Is Not Random Mixing
If the tutor simply shuffles every topic in the syllabus together, difficulty rises.
Information may not.
Good interference work chooses competitors deliberately.
Two methods learners commonly confuse.
Two concepts sharing vocabulary.
Two answer forms with similar surfaces.
The goal is not randomness.
The goal is discrimination under controlled competition.
11. Teach the Discriminating Cue
Once the competition is visible, ask:
What feature should trigger A, and what feature should trigger B?
This becomes the selection rule.
Percentage change:
measure change relative to the starting/reference quantity.
Percentage error:
measure error relative to the accepted or actual value.
Product Rule:
the outermost structure is multiplication of functions.
Chain Rule:
one function is nested inside another.
The student should eventually learn the trigger before the procedure.
12. Ask for a Boundary Sentence
A powerful student task is to complete:
I use A when ________. I use B when ________. The easiest way to tell them apart is ________.
This forces the learner to compress the distinction into a usable cue.
If the sentence is vague, the distinction is probably vague.
If the learner can produce a precise boundary and then apply it across fresh mixed examples, the interference problem is shrinking.
13. Ask for a Near-Miss
Do not ask only for correct examples.
Ask the learner to design a case where Rule A is tempting but wrong.
This reveals whether the boundary is understood.
Maya creates two percentage questions with almost identical numbers but different reference quantities.
Hana writes two sentences where pronoun reference is clear in one and ambiguous in the other.
Jia Jun describes two water-process situations with opposite directions of change.
Ethan writes one function requiring Chain Rule only and another requiring Product Rule plus Chain Rule.
Generation turns discrimination into construction.
14. The Interference Test Is Different From the Diagnostic Probe
The Diagnostic Probe is broader.
It can distinguish concept failure from language load, time pressure from knowledge, selection from execution and many other competing causes.
The Interference Test owns one specific hypothesis:
Is another learned response intruding because the two situations are insufficiently discriminated?
The Diagnostic Probe supplies the general method.
The Interference Test supplies the specialised question.
15. The Interference Test Is Different From the Reactivation Cost
The Reactivation Cost asks how much support is required to restore a faded capability.
Interference can create an apparently high Reactivation Cost if the tutor tries to reactivate the old rule in isolation without addressing the competitor.
The old rule works during blocked review.
Then fails again as soon as the new rule returns.
That is not stable reactivation.
It is unresolved competition.
16. The Interference Test Is Different From the Transfer Gate
The Transfer Gate tests portability into changed contexts.
The Interference Test deliberately introduces a competing neighbour.
Transfer asks:
“Can the skill travel?”
Interference asks:
“Can the skill remain correctly selected when a similar alternative is also available?”
That is a harder and more specific test.
17. The Interference Test Is Different From Performance Reliability
Performance Reliability asks whether a capability can be repeated consistently.
Interference asks what happens when a competing rule is activated nearby.
A skill can be perfectly reliable in isolation and unreliable in competition.
That distinction is common in school examinations, where topics are rarely announced politely before each question.
18. Primary 1: Similar Letter and Sound Patterns Can Compete
Young learners often meet patterns that look or sound similar.
When a newly taught pattern appears, an older one may become less reliable for a while.
Do not immediately conclude that the child has lost the earlier learning.
Use a small contrast.
Two words.
Point to the distinguishing letters.
Sound them.
Switch order.
Return later.
The task is discrimination, not punishment through volume.
19. Primary 2: New Spelling Patterns Can Distort Old Ones
Ethan learns a new spelling convention.
Suddenly an older family becomes unstable.
Test both families side by side.
Ask what visual or sound feature changes the choice.
If each is correct alone but mixed work fails, the old knowledge is not the only problem.
The child needs a cleaner boundary.
20. Primary 3 and Primary 4: Science Vocabulary Creates High-Interference Zones
Science contains many concept pairs that share language.
Evaporation and condensation.
Heat and temperature.
Mass and weight.
Observation and inference.
Students may know each definition separately yet confuse them when a new experiment contains overlapping surface cues.
Contrast the cases.
Ask what changes direction, mechanism or measurement.
The tutor is teaching conceptual boundaries.
21. Primary 5: Interference Matters Before PSLE Volume Increases
By Primary 5, learners have accumulated many similar methods.
Ratio.
Fractions.
Percentage.
Rate.
Area and perimeter.
Vocabulary with near-synonyms.
Science explanations with related processes.
Blocked topical practice can hide selection problems.
Primary 5 is a good time to introduce carefully chosen mixed contrasts before PSLE practice becomes dense.
22. Primary 6: Full Papers Are Interference Environments
A full PSLE-style paper does not announce:
“This is a ratio question.”
“This is an inference question.”
“This is evaporation.”
Methods compete.
Representations compete.
Old habits compete with new ones.
This is why examination practice should include discrimination, not merely more quantity.
If an old skill fails only after a similar new topic was introduced, isolate the pair before declaring broad regression.
23. Secondary 1: New Notation Can Interfere With Old Representations
Secondary 1 changes how familiar ideas are represented.
A learner may know the relationship but map new symbols onto an old rule incorrectly.
Test the concept in the familiar representation.
Then in the new representation.
Then compare them side by side.
If the old form works and the new form distorts it, the bridge is the learning job.
Do not rebuild the entire Primary foundation.
24. Secondary 2: Similar Algebraic Procedures Can Compete
As algebra expands, several legitimate operations become available.
Expand.
Factorise.
Rearrange.
Simplify.
Substitute.
A learner can know every procedure and still select an expensive or wrong one.
Interference here is often procedural competition.
Ask:
What feature of the expression makes this operation useful now?
Selection becomes explicit.
25. Secondary 3: Additional Mathematics Is a Dense Interference Environment
Secondary 3 Additional Mathematics creates a perfect interference laboratory.
Product Rule.
Chain Rule.
Quotient structures.
Trigonometric identities.
Logarithmic laws.
Coordinate forms.
Several methods can be mathematically valid in nearby-looking situations.
Blocked chapters make each rule look easy.
Mixed questions reveal whether the learner can choose.
The tutor should teach rule boundaries before assuming every mixed error is weak calculus.
26. Secondary 4: Interference Appears Under Speed
Under relaxed practice, Ethan distinguishes two methods correctly.
Under full-paper speed, the first familiar method wins.
This is important.
The knowledge boundary may exist but not yet dominate under pressure.
The Interference Test should therefore be repeated at realistic speed after discrimination is learned.
Examinations reward the rule that can be selected fast enough, not merely explained eventually.
27. JC: Interference Can Become Strategy Competition
At JC, the learner may have many legitimate strategies.
The problem is no longer remembering one formula.
It is selecting the representation or method that best fits the structure.
Old successful habits can become proactive interference when they are over-applied to newer problem forms.
Advanced tuition therefore asks not only:
“Can you use this method?”
but:
“What evidence tells you this method should beat the alternatives here?”
28. English: Near-Synonyms Are Interference Problems
Hana knows reluctant.
She learns hesitant.
Now she uses them interchangeably.
The issue is not that she has forgotten both words.
The semantic boundary is weak.
Contrast:
reluctance involves unwillingness.
hesitation involves uncertainty or delay.
Then write two sentences where only one word fits naturally.
Vocabulary precision is often boundary learning.
29. English: Writing Formats Can Interfere
A learner becomes competent in one situational-writing form.
Then another form is taught.
Audience, tone, opening or closing conventions begin to bleed across tasks.
Reteaching both formats separately may not solve the problem.
Put them side by side.
Ask what audience relationship changes the choice.
The learner needs a decision boundary before another model answer.
30. English: Grammar Rules Can Compete
Grammar learning contains many overlapping cues.
Subject–verb agreement.
Tense consistency.
Pronoun reference.
Relative clauses.
When a new rule appears, a learner may overgeneralise it.
The Interference Test asks:
Does the learner still perform each rule in isolation?
Which sentence feature should trigger each?
Can the learner identify which rule is irrelevant?
Knowing when not to apply a rule is part of knowing the rule.
31. Mathematics: Formula Competition
Formulae that share variables or similar shapes can compete.
The student recognises the topic.
Retrieves a related formula.
Substitutes confidently.
Everything looks fluent.
Except the governing relationship is wrong.
Do not solve formula competition only by making the student copy each formula ten times.
Ask what condition makes one formula appropriate.
Then alternate cases.
32. Mathematics: Representation Competition
Sometimes the competition is not between formulae.
It is between representations.
Bar model.
Equation.
Ratio table.
Graph.
A learner may force the familiar representation onto a problem where another is cleaner.
The old successful representation is now interfering with flexible choice.
The learning job is not to ban the old method.
It is to add a selection rule.
33. Science: Concept Pairs Need Contrastive Teaching
Jia Jun knows evaporation.
Then learns boiling.
Both involve liquid becoming gas.
Now the shared feature is strong.
The boundary must become stronger too.
Where does the process occur?
At what temperatures can it occur?
What observable evidence differs?
Contrast protects both concepts.
34. Science: Evidence Rules Can Also Interfere
A student learns:
“If two variables change together, they are related.”
Later learns:
“A fair test changes one independent variable.”
Now the learner may apply the fair-test rule to observational data where experimental control was never available.
Or treat correlation as proof of causation.
The answer is not more slogans.
The answer is to distinguish what claim each evidence pattern licenses.
35. Additional Mathematics: Product Rule and Chain Rule
Ethan can perform both rules.
Blocked Product Rule set:
excellent.
Blocked Chain Rule set:
excellent.
Mixed set:
errors.
The problem is not necessarily either rule.
Ask for structure before differentiation.
“What is the outermost operation?”
“Is there nesting inside it?”
The learner now has a hierarchy:
identify structure first, then select rules.
That is interference repair.
36. Additional Mathematics: Old Algebra Can Interfere With New Structure
Students sometimes manipulate too early.
They see algebra and begin expanding because expansion has been successful historically.
But the new A-Math problem is easier if structure is preserved.
The old habit is not wrong in general.
It is wrong here.
Advanced learning often requires inhibiting a familiar response when a newer structure demands a different one.
37. Interference Can Be Asymmetric
Rule A may interfere strongly with Rule B while Rule B barely affects Rule A.
The learner may have practised A for years.
B is new.
Under pressure, A wins.
Or a vivid new rule may temporarily dominate an older one.
This is why the tutor should test both directions.
Old → new.
New → old.
Do not assume the confusion is perfectly balanced.
38. Retrieval Can Help Reset Context Between Learning Sets
Research reviewed by Bernhard Pastötter and Karl-Heinz Bäuml describes a forward effect of testing: retrieving earlier material can improve learning of subsequently presented information, with release from proactive interference among the proposed mechanisms.
For tuition, the cautious practical implication is useful.
Before launching a similar new rule, briefly retrieve the old one.
Then state the boundary.
Then teach the new rule.
This can make the old state explicit rather than allowing the new learning to arrive into an unexamined memory field.
The claim should remain modest: retrieval has broad evidence for strengthening later accessibility, while the exact mechanisms and best classroom implementation depend on the material and context.
39. Retrieval Should Include Selection, Not Only Recall
If interference is the problem, retrieving both rule definitions separately may not be enough.
The learner must retrieve the selection boundary.
Not:
“What is Product Rule?”
and:
“What is Chain Rule?”
But:
“What structural feature tells you which rule comes first?”
This is retrieval of discrimination.
40. Spacing Can Reveal Interference That Immediate Practice Hides
Immediately after contrastive teaching, the learner remembers the distinction because the tutor just discussed it.
That is useful but incomplete evidence.
Return later.
One week.
Mixed questions.
No labels.
If the distinction survives, confidence rises.
If the rules begin bleeding into each other again, the boundary needs maintenance.
This connects directly to The Maintenance Window.
41. The Interference Matrix
A tutor can use a simple four-cell comparison.
A alone: can the learner execute old rule A?
B alone: can the learner execute new rule B?
A vs B selection: can the learner identify which rule applies before calculating?
A/B mixed execution: can the learner select and execute both across alternating cases?
Interpretation:
- A fails alone → old rule needs reactivation or repair.
- B fails alone → new rule is not yet learned securely.
- A and B pass alone but selection fails → interference/discrimination is strongly implicated.
- Selection passes but execution fails under speed → performance conditions may be the problem.
Four small tests can replace one vague claim:
“She is confused.”
42. The Interference Ladder
A practical repair sequence is:
Level 1 — Re-establish each rule alone.
Confirm both are genuinely available.
Level 2 — Name the distinguishing cue.
What feature separates the cases?
Level 3 — Compare matched pairs.
Use near-neighbours differing on one decisive feature.
Level 4 — Alternate cases.
A, B, A, B without announcing the rule.
Level 5 — Mix with unrelated distractors.
Selection must now occur inside a broader field.
Level 6 — Delay.
Return after the immediate contrast has cooled.
Level 7 — School return.
Observe whether the correct boundary survives authentic work.
This is not a validated scale.
It is an explanatory teaching sequence.
43. Do Not Solve Interference With More Blocked Practice
Maya confuses A and B.
The tutor assigns twenty A questions.
Maya becomes excellent at A.
Then twenty B questions.
Maya becomes excellent at B.
Then a mixed paper arrives.
Confusion returns.
Why?
The training strengthened execution without strengthening selection.
Once individual fluency is adequate, the repair must move to mixed discrimination.
44. Do Not Solve Interference by Deleting the Old Rule
Sometimes adults say:
“Forget the old method. Use this one now.”
That may simplify the lesson.
But if both methods remain valid in different situations, deleting one reduces flexibility.
The better objective is conditional knowledge.
Use A here.
Use B there.
Know why.
Know the boundary.
45. Do Not Introduce Similar Rules Without an Explicit Boundary
Prevention is cheaper than repair.
When teaching a new rule near an old one, begin with:
“This looks similar to what you already know.”
Then:
“Here is what stays the same.”
Then:
“Here is the feature that changes the decision.”
This is contrastive encoding.
The old rule is protected because its territory is named while the new rule is learned.
46. Interference Can Be Productive
Competition is not always bad.
When two strategies are available, the learner has an opportunity to compare them.
Which is more efficient?
Which is more general?
Which exposes structure?
Which is safer under examination time?
What begins as interference can become adaptive expertise if the learner learns to discriminate and choose deliberately.
The goal is not a mind containing only one response.
The goal is a mind that selects among useful responses well.
47. The Priority Queue Should Promote High-Interference Pairs
A learner repeatedly loses marks because two common methods are confused.
That pair deserves attention.
Why?
Recurrence is high.
Transfer value is high.
The repair may be cheap.
One precise boundary can prevent errors across many future questions.
This is exactly the sort of high-value job that belongs near the front of The Priority Queue.
48. The Maintenance Window Should Track Known Competitors
A skill was stable until a similar new topic appeared.
After the interference repair, do not maintain the old rule alone.
Maintain the pair.
One A.
One B.
One explanation of the boundary.
Then widen the interval.
The maintenance object is no longer Rule A.
It is the discrimination between A and B.
49. The Reactivation Cost Should Be Measured With the Competitor Present
A skill reactivates perfectly in isolation.
Then collapses when its neighbour appears.
Do not record:
“reactivation successful.”
Record:
“execution restored alone; selection remains unstable under interference.”
The cost of useful reactivation includes restoring correct selection, not merely isolated procedure.
50. The Evidence Threshold for Interference Should Be Modest but Real
One wrong answer after new learning is not enough to diagnose interference confidently.
But the threshold for a small contrastive test is low.
Run the A/B comparison.
If each rule is correct alone and repeatedly confused together, the evidence becomes much stronger.
The Evidence Threshold still governs how far the response should go.
Small contrastive practice?
Low threshold.
Rewrite the whole tuition plan?
Much higher threshold.
51. The Counterfactual Check Can Ask Whether New Learning Caused the Old Failure
Old skill deteriorates after new topic begins.
Did the new learning cause the change?
The Counterfactual Check reminds us not to overclaim.
Perhaps the timing is coincidental.
But if:
- A was recently stable;
- B was introduced;
- A then began receiving B-type responses;
- A and B both work alone;
- mixed A/B selection fails;
- contrastive training restores both;
the interference explanation becomes considerably more coherent.
Still practical evidence.
Not laboratory proof.
52. The Parent Should Not Call Interference “Confusion” Too Quickly
“She is confused.”
That phrase can become another broad identity label.
Better:
“She can do both rules separately, but she is currently selecting between them unreliably.”
That is smaller.
More testable.
More repairable.
And kinder because it describes a decision problem rather than a child.
53. The Tutor’s Interference Test
- Was the old capability previously stable?
- Did deterioration begin after similar new learning?
- Is the current wrong answer actually a correct response from the competing rule?
- Can the learner execute old Rule A in isolation?
- Can the learner execute new Rule B in isolation?
- Can the learner state the difference between A and B?
- Can the learner identify the discriminating cue before solving?
- Can the learner alternate A and B without headings?
- Does performance remain stable after delay?
- Does the distinction survive examination speed?
- Does the pair need maintenance together?
This turns “confusion” into an observable learning problem.
54. The Student’s Interference Test
An older learner can ask:
- What two rules am I mixing?
- Can I still do each one separately?
- What feature tells me which one applies?
- Can I explain why the other rule is wrong here?
- Can I create one example for each?
- Can I create a near-miss?
- Can I alternate them without chapter labels?
- Can I still choose correctly next week?
This is far more useful than writing:
“Revise both topics.”
55. The Parent’s Interference Test
Parents do not need to teach the contrast.
They can notice timing.
“This started after the new topic.”
They can preserve examples.
“She used the new formula in the old question twice.”
Then send the evidence to the tutor.
No kitchen-table re-teaching war is required.
The family role is observation and routing.
56. The Long Return: Experts Manage Interference Too
Adults also learn systems that resemble older systems.
A programmer learns a new language and writes old syntax.
A driver changes countries and instinctively follows familiar road habits.
A professional adopts a new software workflow and reaches for the old shortcut.
An expert updates a guideline and must suppress the previous recommendation when the conditions now differ.
Competence does not mean old learning disappears when new learning arrives.
Competence means the system can choose appropriately among overlapping memories.
School can begin teaching that architecture early.
57. Back to Maya
The tutor gives Maya four questions.
Percentage change.
Percentage error.
Percentage change.
Percentage error.
No headings.
Before each calculation, Maya must write only one thing:
reference = ______ because ______
At first she hesitates.
Then the distinction sharpens.
Starting value for percentage change.
Accepted or actual value for percentage error.
She gets all four right.
The tutor does not celebrate yet.
He adds two unrelated ratio questions between them.
Still right.
Next week, the pair will return cold.
If school provides a good example first, that may count as the next maintenance event.
The old skill was not simply forgotten.
The new skill was not simply bad teaching.
Two useful rules entered the same neighbourhood.
Maya needed a clearer road sign.
That is what the Interference Test found.
The Interference Test in One Page
1. Do not equate every old-skill failure with forgetting.
New similar learning can compete with old learning.
2. Look at timing.
A sudden decline after a related new topic is worth testing.
3. Look for substituted rules.
A coherent wrong-but-related response is a clue.
4. Test A alone and B alone.
Confirm whether each rule still exists independently.
5. Then test A versus B.
Selection is a separate capability from execution.
6. Teach the discriminating cue.
What feature tells the learner which rule applies?
7. Use matched contrasts.
Near-neighbours expose the boundary efficiently.
8. Ask why the competing rule is wrong.
Boundary knowledge matters as much as correct procedure.
9. Move from blocked to mixed practice.
Once execution is stable, discrimination needs practice.
10. Do not randomise blindly.
Choose competitors intentionally.
11. Test both directions.
Old learning can disrupt new learning and new learning can disrupt access to old learning.
12. Retest after delay.
Immediate contrast can create temporary clarity.
13. Retest under realistic speed.
Pressure can make the familiar competitor win.
14. Maintain the distinction, not merely each rule separately.
The A/B boundary becomes the maintenance object.
15. Reframe “confusion.”
Describe the selection problem precisely enough to repair it.
Where This Fits in eduKatePunggol
- How Tuition Works at eduKatePunggol — the broad tuition relationship.
- How Tuition Works | The Weak Link — identify the first limiting mechanism.
- The Flight Recorder — show that the old skill was stable before the new competitor arrived.
- The Learning Dispatcher — decide where contrastive practice belongs.
- The Exit Ramp — preserve independence once the learner can discriminate without tutor prompts.
- The Priority Queue — promote recurrent high-consequence interference pairs.
- The Intervention Threshold — avoid cueing the rule before the learner has revealed the selection state.
- The Translation Layer — describe the substitution precisely rather than calling it generic confusion.
- The Transfer Gate — test whether each rule survives changed conditions.
- The Diagnostic Probe — distinguish interference from decay, weak knowledge, timing or representation.
- The Evidence Threshold — decide when evidence justifies a larger change.
- The Counterfactual Check — assess whether the new learning plausibly contributed to the old failure.
- The Maintenance Window — maintain known competing rules through spaced contrast.
- The Reactivation Cost — distinguish quick retrieval recovery from unresolved competition.
Batch 04 now has three distinct mechanisms:
- The Maintenance Window — when should stable learning return?
- The Reactivation Cost — how expensive is it to restore after fading?
- The Interference Test — is apparent forgetting actually competition from similar learning?
The maintenance system is becoming more precise:
Leave stable learning alone most of the time → return after a useful interval → if it fails, measure reactivation → if a similar new rule is present, test interference → repair the boundary rather than blindly reteaching the chapter.
Research Foundations
The Interference Test is an explanatory tuition model, not a validated diagnostic instrument. Memory research has long studied proactive and retroactive interference, retrieval competition and the effects of similarity between memories. Classroom learning is more complex than laboratory list-learning, so these mechanisms should be used as hypotheses to test rather than labels to impose.
- Pastötter & Bäuml — Retrieval Practice Enhances New Learning: The Forward Effect of Testing, Frontiers in Psychology (2014). The review discusses evidence that retrieval practice on earlier material can improve subsequent learning and examines release from proactive interference among the theoretical explanations.
- McDermott — Practicing Retrieval Facilitates Learning, Annual Review of Psychology (2021). This review summarises broad evidence that retrieving newly learned information supports later accessibility across many materials and learners.
- Rohrer, Dedrick & Stershic — Interleaved Practice Improves Mathematics Learning. The classroom study compared blocked and interleaved Mathematics practice and found an advantage for interleaving on a delayed test; the authors discuss strategy discrimination as an important mechanism.
- Roediger & Butler — The Critical Role of Retrieval Practice in Long-Term Retention, Trends in Cognitive Sciences (2011). The review discusses retrieval practice as a powerful learning tool with implications for durable and flexible access to knowledge.
- Beyond Retrieval Competition: Asymmetric Effects of Retroactive and Proactive Interference in Associative Memory (2025). This recent laboratory work is a useful reminder that proactive and retroactive interference need not be mirror images and that interference mechanisms may include more than simple retrieval competition.
The practical conclusion is simple.
When an old skill becomes unreliable after a similar new one arrives, do not immediately reopen the old chapter.
Put the two neighbours side by side.
Ask what makes them different.
Ask what cue should trigger each.
Then mix them.
Delay them.
Return them to school.
The goal is not merely to remember both rules.
The goal is to know which rule belongs to which world.
