Maya gets the question wrong.

That is the easy part.

The difficult part is deciding why.

The Mathematics question asks for a percentage decrease.

Maya subtracts correctly.

Then divides by the new value instead of the original value.

The tutor can tell at least two stories.

Story A: Maya does not understand what percentage change means.

Story B: Maya understands the idea but loses track of the reference quantity when the wording becomes dense.

Both stories fit the wrong answer.

If the tutor chooses Story A, the next twenty minutes may become a full reteaching of percentage change.

If Story B is true, that teaching is unnecessarily broad.

If the tutor chooses Story B and Maya actually lacks the concept, a tiny reminder about the “original amount” will be too weak.

One error.

Two plausible causes.

Which one should the tutor believe?

Neither yet.

Ask a better question.

The tutor writes:

A price falls from $80 to $60. Without calculating, which amount should the percentage decrease be compared with: $80 or $60? Explain why.

Maya answers immediately:

“$80, because we are measuring how much it changed from where it started.”

Now the tutor knows more.

The core meaning is present.

The original error is more likely to involve maintaining the reference quantity through a multi-step question than complete conceptual absence.

The new question did not simply test Maya again.

It separated two explanations.

That is a Diagnostic Probe.

When one performance can be explained by two different causes, design the smallest fresh task that makes those causes predict different responses.

This is one of the most valuable things a small tuition room can do.

Do not teach the first plausible story.

Make the stories compete.

Quick Read: The Diagnostic Probe in One Sentence

A Diagnostic Probe is a deliberately chosen question, example or contrast whose answer tells the tutor which of two or more plausible learning explanations is better supported.

The probe is not a quiz for its own sake.

It exists because a decision must be made.

Reteach or practise?

Vocabulary or comprehension?

Concept or calculation?

Knowledge or confidence?

Method selection or execution?

Time pressure or missing understanding?

The probe should make those possibilities behave differently.

1. The Wrong Answer Is Usually Underdetermined

A wrong answer tells us that something went wrong.

It rarely tells us everything we want to know about why.

A blank can mean no knowledge, no time, no confidence or no strategy.

A wrong calculation can mean a conceptual misconception, a copied number, a weak arithmetic fact, a notation problem or an efficient method executed badly.

A weak English answer can mean poor comprehension, thin vocabulary, an unsupported inference, misunderstanding of the command word or failure to express an idea already understood.

A short Science explanation can mean missing concept knowledge, missing causal linkage, weak writing or simple haste.

This is why the Translation Layer comes first.

It turns the broad signal into an accurate description of the observed behaviour.

Then the Diagnostic Probe asks:

What is the smallest additional evidence we need before choosing between the remaining explanations?

2. The Probe Is Different From the Weak Link

How Tuition Works | The Weak Link asks which mechanism is failing first in a learning chain.

The Diagnostic Probe is one instrument used when that mechanism is uncertain.

For example:

Maya fails a word problem.

Possible weak link:

reading the relationship.

Possible weak link:

solving the resulting equation.

Probe:

Ask Maya to form the equation but not solve it.

If she forms it correctly, representation is less likely to be the first weak link.

If she cannot, the solving algorithm was never the main issue.

The probe narrows.

The Weak Link names the resulting target.

3. The Probe Is Different From the Priority Queue

The Priority Queue decides which known learning problem deserves the next minute.

The Diagnostic Probe is what the tutor may use before the queue is confident enough to rank the problem.

Sometimes one minute spent probing prevents thirty minutes spent repairing the wrong thing.

This means diagnosis itself has opportunity cost.

The probe should be small.

Not another full test.

Not ten random questions.

One discriminating question if one will do.

4. The Probe Is Different From the Transfer Gate

The Transfer Gate asks whether a learned capability survives changed conditions.

A Diagnostic Probe asks what caused a specific success or failure.

Transfer Gate result:

Hana fails after the topic label is removed.

Possible cause A:

she cannot recognise the strategy without the label.

Possible cause B:

the new passage is simply too difficult linguistically.

Probe:

give an easier passage with the label removed.

If she still fails to activate the strategy, cue dependency becomes more plausible.

Transfer failure creates the question.

The probe separates the explanations.

5. The Probe Is Different From the Intervention Threshold

The Intervention Threshold asks when the tutor should step into a live attempt.

The Diagnostic Probe asks what task the tutor should put in front of the learner when more evidence is needed.

The two mechanisms work together.

Design the probe.

Then allow enough independent response for the probe to remain diagnostic.

If the tutor hints too early, the probe is contaminated.

6. Good Formative Assessment Begins With the Decision

The U.S. Institute of Education Sciences makes a useful assessment point: begin with the decision you need to make, then choose evidence suited to that decision.

A broad assessment may tell you whether a student is on track.

It may not tell you what to do in the next ten minutes.

A Diagnostic Probe is explicitly local.

The tutor asks:

What decision am I unable to make because two explanations still fit the evidence?

Then designs evidence for that decision.

7. Hinge Questions Show the Power of Discriminating Answers

Dylan Wiliam’s work on hinge questions offers a strong adjacent idea.

A good hinge question is not merely a quick multiple-choice check.

Its response options are designed so different answers reveal different misunderstandings and help the teacher decide whether to move on, reteach or adapt.

EEF’s current diagnostic-assessment materials make the same practical point: carefully designed wrong answers can be built around known misconceptions so that responses provide information about what pupils understand.

The Diagnostic Probe uses that logic at the scale of one learner.

Instead of asking:

“Can Maya answer another percentage question?”

ask:

“What one percentage question would make conceptual misunderstanding and reference-tracking failure produce different behaviour?”

8. The Probe Needs Competing Explanations

Do not design the probe until you can state at least two plausible explanations.

Bad diagnostic thinking:

“Maya is weak at percentages. Let me test percentages.”

Better:

“Maya may not understand percentage change, or she may understand it but lose the reference whole when the question becomes multi-step.”

Now the tutor can design a discriminating question.

The quality of the probe depends on the quality of the alternatives.

9. One Cause Must Predict Something the Other Does Not

This is the heart of the mechanism.

If two explanations predict exactly the same answer to your new question, the question is not discriminating.

Suppose:

A = does not understand the meaning of percentage decrease.

B = understands the meaning but loses track of the reference amount in dense working.

Ask an easy conceptual question with no calculation.

A predicts likely failure.

B predicts likely success.

Now the answers diverge.

That is a probe.

10. Change One Variable Where Possible

Good diagnostic probes resemble controlled comparisons.

Keep most of the task stable.

Change the feature that separates the hypotheses.

If you suspect language rather than Mathematics, simplify the language while preserving the mathematics.

If you suspect method selection rather than execution, provide the method and test execution.

If you suspect memory rather than understanding, give a recognition version before a recall version.

If you suspect time pressure rather than knowledge, remove the timer.

If you suspect writing rather than Science understanding, ask for an oral explanation.

The probe changes one informative feature and watches what happens.

11. The Simplification Probe

This is one of the most useful probes in tuition.

Keep the concept.

Reduce surface complexity.

Hana fails a dense comprehension inference.

Possible cause A:

she does not understand inference.

Possible cause B:

sentence complexity overloaded her reading.

Probe:

use a short, linguistically simple passage requiring the same inferential relationship.

Success points toward language load.

Failure keeps the inference mechanism in contention.

12. The Representation Probe

Keep the relationship.

Change how it is represented.

Maya fails a word problem involving simultaneous equations.

Give her the equations directly.

If she solves them, equation-solving is less likely to be the problem.

Then ask her to form equations from a simpler word relationship.

The two probes can isolate whether language-to-equation representation is the boundary.

The tutor has avoided reteaching an algorithm Maya already owns.

13. The Oral-Written Probe

Jia Jun writes a weak Science explanation.

Possible cause A:

the concept is missing.

Possible cause B:

the concept is understood but not being externalised clearly in writing.

Probe:

“Explain it to me without writing.”

If Jia Jun gives a complete causal explanation orally, the tuition job changes.

Do not reteach the Science concept first.

Train translation from oral reasoning into written causal structure.

14. The Recognition-Recall Probe

Ethan says he does not know the formula.

Possible cause A:

the formula was never learned securely.

Possible cause B:

the formula is available in recognition but weak in free recall.

Probe:

show four plausible formulas and ask which one applies and why.

If Ethan recognises and explains the correct relationship, the conceptual representation may be present while retrieval strength remains weak.

That suggests retrieval practice rather than full concept teaching.

15. The Timed-Untimed Probe

Maya fails four questions in the final third of a paper.

Possible cause A:

the underlying knowledge is weak.

Possible cause B:

the knowledge is stable but performance collapses under time and fatigue.

Probe:

reattempt comparable questions untimed while rested.

Immediate success does not prove time is the only cause.

But it changes the probability enough to justify a second performance probe.

Then reproduce the time condition deliberately.

Diagnosis may require a sequence, not one heroic question.

16. The Label-On, Label-Off Probe

Maya can differentiate when the worksheet says Chain Rule.

She fails in a mixed paper.

Possible cause A:

she cannot execute the Chain Rule.

Possible cause B:

she can execute it but cannot recognise when it is needed.

Probe:

give one unlabelled nested function and ask only:

“Which differentiation rules are present? Do not differentiate yet.”

The probe removes execution and isolates recognition.

17. The Method-Given Probe

A student gets a problem wrong.

Was the method wrong or the execution?

Give the method.

“Use simultaneous equations.”

Now observe.

If the learner succeeds, method selection remains the likely weak point.

If the learner still fails, the problem sits lower in the execution chain.

This simple probe can save substantial reteaching.

18. The Execution-Removed Probe

The reverse is equally useful.

Remove calculation and ask only for the decision.

“Which method would you use?”

“Which graph feature matters?”

“Which paragraph contains the strongest evidence?”

“Which variable is controlled?”

If decision-making is correct when execution is removed, the tutor has separated a higher-order choice from a lower-level procedure.

19. The Contrast Probe

Sometimes one example is not enough.

Use a pair.

Question A should trigger the target rule.

Question B should look similar but require a different rule.

Now ask:

“What makes these two questions different?”

This is powerful because misconceptions often survive isolated examples but collapse under contrast.

For Mathematics:

percentage of original value versus percentage of new value.

For English:

supported inference versus plausible but unsupported inference.

For Science:

correlation versus evidence for a causal relationship.

Contrast makes the discriminating feature visible.

20. The Counterexample Probe

A student states a rule that is too broad.

Do not immediately lecture about exceptions.

Give one case that should break the rule.

If the learner notices the contradiction, the concept may be flexible.

If the learner forces the old rule onto the counterexample, the misconception has become clearer.

This is where the Sparring Partner can become diagnostic.

Resistance does not merely make thinking harder.

It can reveal exactly which rule the learner believes.

21. The Explain-Your-Choice Probe

A correct answer can still be diagnostically ambiguous.

Maya selects the correct option.

Was it knowledge or luck?

Ask why.

If the reasoning is sound, confidence rises.

If the reasoning is wrong but happens to land on the correct option, the probe has prevented a false positive.

Wiliam’s writing on hinge questions stresses this problem: a diagnostic question should be designed so that teachers do not mistake a right answer produced for the wrong reason as secure understanding.

22. The Commit-Before-Hint Probe

A student says:

“I don’t know.”

Possible cause A:

knowledge truly missing.

Possible cause B:

knowledge present but confidence too low to commit.

Probe:

“If you had to choose, which answer would you choose? Give me one reason.”

If the learner chooses correctly and reasons well, the tutor should not reteach the content reflexively.

The learning job may be commitment and confidence calibration.

23. The First-Step Probe

Full problems contain too many stages.

Ask only for the first one.

“What would you do first?”

This is especially powerful in Mathematics and Science.

The learner does not have to finish the computation.

The tutor sees whether the problem has been classified appropriately.

A correct first move with later failure points downstream.

A wrong first move points upstream.

24. The Last-Step Probe

Sometimes the learner reaches a correct intermediate result but loses marks at the end.

Is the issue calculation?

Interpretation?

Units?

Answer form?

Give the correct penultimate value and ask only for the final response.

This isolates whether the learner can convert mathematics into the requested answer form.

Again, remove irrelevant difficulty to expose one decision.

25. The Error-Choice Probe

Show two wrong solutions.

Ask:

“Which student made the earlier mistake?”

Or:

“Which answer is wrong for a more fundamental reason?”

This can reveal whether the learner understands the governing principle well enough to diagnose someone else’s work.

Sometimes students can recognise an error before they can avoid producing it.

That distinction itself is diagnostic.

26. The Example-Generation Probe

A student says they understand the rule.

Ask them to generate an example.

“Give me a pair of quantities where the percentage increase is 25%.”

“Write a sentence where the pronoun reference is ambiguous.”

“Design an experiment where two variables change and the causal conclusion would therefore be unsafe.”

Generation tests a different direction of knowledge.

A learner who can recognise but cannot generate may possess a narrower representation of the concept.

27. The Boundary Probe

A learner uses a rule successfully.

Does the learner know when it stops applying?

Give a near-neighbour where the rule is tempting but wrong.

This helps distinguish genuine structural understanding from overgeneralised procedure.

The Boundary Probe is especially important for strong students because fluent success can hide rules that have become too broad.

28. The Delay Probe

Immediate performance is strong.

Is the problem solved?

Wait.

Ask again next week.

Delay can separate temporary activation from durable retrieval.

This is not always a one-question probe in the same lesson.

Sometimes the discriminating variable is time.

The Flight Recorder is what makes such probes usable because it remembers the earlier state.

29. Primary 1: Probe Gently

Young learners should not feel as if every mistake has become an interrogation.

A Primary 1 probe can be playful and tiny.

Maya reads ship as shop.

Does she confuse the medial vowel or guess from first and last letters?

Probe with:

ship, shop, shut.

Ask her to point while sounding through the middle.

One minute can show whether the vowel representation or whole-word guessing deserves attention.

30. Primary 2: Probe Reading Fluency Carefully

Ethan reads slowly.

Possible cause A:

decoding remains effortful.

Possible cause B:

decoding is adequate but unfamiliar vocabulary is slowing interpretation.

Probe with two short passages matched in sentence structure but different in vocabulary familiarity.

If speed changes dramatically only with vocabulary load, the teaching job changes.

Do not call all slow reading the same thing.

31. Primary 3 and Primary 4: Probe Science Concepts Through Prediction

Jia Jun memorises the phrase:

“More light helps plants grow.”

Does he understand the relationship or merely the phrase?

Give a new case.

Two identical plants receive the same water, soil and temperature, but one receives almost no light.

Ask him to predict and explain.

Then alter one more variable and ask whether the conclusion is still safe.

The probe separates memorised association from evidence-aware reasoning.

32. Primary 5: Probe Before the PSLE Volume Arrives

Primary 5 is a dangerous year for broad labels.

Weak comprehension.

Weak problem sums.

Weak Science.

Once full-paper practice accelerates, these labels can create large quantities of undirected work.

Use Primary 5 runway to probe.

Which inference?

Which representation?

Which Science relationship?

Which writing decision?

Better diagnosis now makes Primary 6 practice more selective.

33. Primary 6: Probe Fast, Then Train

Closer to PSLE, diagnostic elegance must respect time.

The tutor cannot spend half the lesson investigating every mark.

Choose high-consequence ambiguity.

If a recurring error could come from either concept failure or time pressure, one quick timed–untimed contrast can change the entire revision plan.

If a comprehension loss could come from vocabulary or inference, one simpler passage can prevent weeks of generic comprehension practice.

Probe where the decision matters.

34. Secondary 1: Probe Transition Before Calling It Foundation Failure

A Secondary 1 student suddenly struggles.

Possible cause A:

Primary foundation was weak.

Possible cause B:

the learner understands the old relationship but is not yet reading the new notation or task format fluently.

Probe by expressing the same concept in a familiar Primary representation and the new Secondary representation.

If the old form is strong and the new form fails, the bridge is the job.

Do not rebuild what is still standing.

35. Secondary 2: Probe Self-Management vs Subject Weakness

A student’s marks begin falling across several subjects.

Possible explanation:

several subject foundations are suddenly weak.

Another explanation:

the student’s planning and homework system has broken as workload increased.

The probe may not be a subject question at all.

Compare what happens when one task is attempted in a structured supervised block versus an ordinary unstructured week.

If subject performance recovers dramatically under structure, the dispatcher should not route every problem to more subject teaching.

36. Secondary 3: Probe the Old Algebra Inside New A-Math

Maya gets an Additional Mathematics question wrong.

Possible cause A:

the new A-Math concept is not understood.

Possible cause B:

the concept is understood but old algebra collapses inside it.

Probe by giving the A-Math reasoning with the algebraic manipulation simplified.

If Maya explains the new concept correctly, return to algebra.

Do not reteach calculus to solve algebra.

37. Secondary 4: Probe Before Spending the Last Weeks

Near final examinations, every large intervention has a cost.

A student loses ten marks in a paper.

Before allocating the next week, separate:

• knowledge missing;
• method not recognised;
• execution unstable;
• time exhausted;
• checking failed;
• one-off noise.

One or two strong probes can change the Priority Queue dramatically.

Late-stage tuition needs high information per minute.

38. JC: Probe Representation Before Teaching More Content

JC Mathematics often contains problems where students possess the necessary concepts but choose expensive representations.

A learner stalls in a calculus question.

Possible cause:

calculus concept missing.

Alternative:

representation creates algebraic complexity that hides a known calculus path.

Probe by changing representation while preserving the underlying problem.

If the student immediately sees the calculus, the next tuition job is representational choice.

Advanced tuition often becomes diagnosis of expensive choices rather than delivery of more formulas.

39. English: Vocabulary or Inference?

Hana answers an inference question wrongly.

She may not understand the inference relationship.

Or one key word may have distorted the passage.

Probe:

give the meaning of the key word explicitly and ask her to answer again.

If she now reasons correctly from evidence, vocabulary access was a major contributor.

If she still overclaims, inference discipline remains the more likely job.

40. English: Comprehension or Expression?

Hana writes a weak answer.

Ask her to explain verbally.

If the oral answer is equally weak, comprehension remains implicated.

If the oral answer is strong but the written one collapses, the bottleneck is closer to expression, answer formulation or language precision.

One modal shift can save a large amount of misdirected comprehension practice.

41. English: Planning or Sentence Craft?

A composition is weak.

Does the writer lack language?

Or did the story architecture fail before sentences were written?

Probe by asking Hana to plan a new prompt without writing the composition.

If the plan itself has no viable turning point, sentence-level tuition is not the first job.

If the plan is strong but the draft becomes flat, move downstream to execution.

42. Mathematics: Concept or Arithmetic?

Maya gets a fraction problem wrong.

Possible cause A:

fraction relationship misunderstood.

Possible cause B:

multiplication fact error.

Probe by replacing awkward numbers with simple ones while preserving the same fraction relationship.

If the reasoning becomes correct, arithmetic fluency may be the more immediate bottleneck.

If the same conceptual error remains, the fraction model needs repair.

43. Mathematics: Reading or Representation?

Maya cannot solve a word problem.

Read it aloud to her.

If she can now represent it, reading fluency or sentence parsing may have been carrying part of the failure.

If she still cannot map the quantities, the problem is more likely mathematical representation.

The tutor should not solve a language problem with more calculation.

44. Mathematics: Method Selection or Execution?

This is perhaps the classic diagnostic split.

Give the method name.

If the learner succeeds, train recognition and selection.

If the learner fails, descend into execution.

Then probe execution further.

Is the formula recalled?

Is substitution correct?

Is algebra stable?

Diagnosis is recursive.

Each answer can generate the next discriminating question.

45. Science: Knowledge or Evidence Reading?

Jia Jun gives the wrong conclusion from a graph.

Possible cause:

the Science concept is missing.

Alternative:

the learner understands the concept but misreads the axes or pattern.

Probe with the same graph stripped of Science context.

“As x increases, what happens to y?”

If the graphical relationship is still misread, data interpretation is implicated.

If the graph is read correctly but the Science conclusion remains wrong, return to the concept.

46. Science: Concept or Causal Writing?

Ask orally.

Then ask for a diagram.

Then ask for writing.

If the concept survives speech and diagram but breaks in writing, the tuition job is not “learn the topic again.”

It is converting a known causal model into a markable explanation.

Mode changes are powerful probes because they preserve the concept while changing the output demand.

47. Additional Mathematics: Chain Rule or Algebra?

Maya differentiates incorrectly.

Give her:

y = (3x + 1)⁵

If she differentiates correctly, the Chain Rule exists.

Then give:

y = x(3x + 1)⁵

If the second fails, the problem may be hierarchy when Product Rule and Chain Rule interact.

Then simplify the algebra inside the product.

If hierarchy remains correct but simplification fails later, separate structural recognition from algebraic execution.

One broad label—“weak differentiation”—has become a sequence of narrow probes.

48. Examination Performance: Knowledge or Time?

Returned paper:

last five questions incomplete.

Do not automatically conclude:

“Weak final topics.”

Probe the same skills untimed.

If strong, reproduce a timed section.

Then inspect where time is lost.

The first probe removes one explanation.

The second finds the performance mechanism.

Good diagnosis often works like narrowing a search tree.

49. Examination Performance: Carelessness or Checking Failure?

The school comment says:

“Careless mistakes.”

The Translation Layer has already made the errors specific.

Now probe whether the learner can detect them after completion.

If Maya immediately finds the error when told “there is one mistake here,” the checking mechanism may exist but fail to self-trigger.

If she cannot find it even when alerted, the underlying checking knowledge may be weak.

Those are different jobs.

50. Confidence: Does the Student Know More Than They Think?

Hana says “I don’t know” often.

The tutor can translate that signal and then probe.

Forced choice.

Confidence rating.

Reason.

If Hana repeatedly chooses correctly while reporting low confidence, the learning problem is not identical to knowledge absence.

Then calibration becomes relevant without the Diagnostic Probe re-owning that broader topic.

51. Strong Students Need Diagnostic Probes Too

Ethan gets the answer right.

There may still be two explanations.

A:

he recognises the structure deeply.

B:

he recognises a familiar surface pattern.

Probe with a near-neighbour or counterexample.

Strong learners often require harder probes because ordinary questions no longer discriminate.

A question everyone can answer tells you little about the boundary of expertise.

52. Struggling Students Need Smaller Probes

A diagnostic task should not become another overwhelming test.

For a struggling learner, reduce everything except the feature you need to inspect.

One sentence.

One diagram.

One pair of numbers.

One choice between two representations.

The purpose is information, not endurance.

A tiny successful probe can reveal intact knowledge hidden by a large failed task.

53. The Probe Should Be Cheap

Diagnosis can expand endlessly if the tutor is not careful.

Another question.

Another test.

Another worksheet.

The student spends the whole lesson being investigated and never receives teaching.

A good probe has high information value relative to time.

If a thirty-second question separates two explanations, do not administer a thirty-minute test.

Diagnostic efficiency protects lesson time.

54. The Probe Should Be Consequential

Do not probe merely because ambiguity is intellectually interesting.

Probe when the answer changes what you would do.

If both explanations lead to the same immediate teaching move, the distinction may not matter yet.

If one explanation leads to reteaching and the other to timed practice, it matters.

If one leads to vocabulary work and the other to inference training, it matters.

If one belongs in tuition and the other belongs in sleep or school clarification, it matters enormously.

The best probes sit at decision forks.

55. The Probe Should Not Telegraph the Answer

A tutor wants to know whether Maya recognises the reference whole.

Bad probe:

“Remember, for percentage decrease we divide by the original amount. So which amount do you divide by?”

The answer is now almost embedded in the question.

Better:

“The price falls from $80 to $60. Which amount should the decrease be compared with, and why?”

A diagnostic question must preserve the decision it claims to measure.

56. The Probe Should Not Change Too Many Things

Maya fails a percentage question.

The tutor gives another question with harder vocabulary, a graph, decimals and a different percentage relationship.

Maya fails again.

What did we learn?

Very little.

Too many variables changed.

A strong probe is intentionally boring in every dimension except the one being tested.

57. The Probe Should Include Plausible Wrong Answers When Multiple Choice Helps

Multiple choice can be diagnostically powerful when distractors represent specific misconceptions rather than random wrong numbers.

EEF’s diagnostic-assessment guidance highlights exactly this principle for hinge questions.

A distractor should mean something.

For percentage change:

• divide by original amount;
• divide by new amount;
• divide by the difference;
• choose the larger number regardless of direction.

Each answer can correspond to a different mental model.

The wrong option becomes evidence.

58. The Probe Should Sometimes Ask Everyone

A three-student tuition group creates a useful variation on classroom hinge questions.

Ask all three students to commit before discussion.

Mini whiteboards.

Folded paper.

Simultaneous hand signals.

Why?

If Ethan answers first, Maya may copy the logic before the tutor sees her original state.

Commit first.

Then discuss.

This preserves diagnostic visibility while still benefiting from peer reasoning afterward.

59. Three Students Can Produce Three Different Diagnoses From One Probe

The tutor asks one percentage probe.

Maya chooses the original amount and explains correctly.

Ethan chooses correctly but cannot explain why.

Jia Jun chooses the new amount because it is “the final value.”

Same thirty seconds.

Three learner states.

Maya may need complex-context practice.

Ethan may need conceptual articulation.

Jia Jun needs the reference relationship rebuilt.

This is where small-group tuition can use one shared question without pretending the students therefore need one shared intervention.

60. The Probe Can Protect Against Tutor Bias

Tutors form expectations.

Maya is usually careless with signs.

So when she gets a new question wrong, the tutor sees signs.

Hana is usually strong at reading.

So when she fails a passage, the tutor assumes vocabulary.

Prior knowledge of the learner is valuable.

It can also become a story that explains everything.

A discriminating probe gives the current evidence a chance to disagree with the tutor’s expectation.

That is one reason to ask rather than assume.

61. The Probe Can Protect Against Parent Labels

“She rushes.”

“He never understands word problems.”

“She just lacks confidence.”

Parents may be right.

They may also be describing a visible pattern whose mechanism is different.

Good tuition respects the observation and probes the explanation.

That protects the parent–tutor relationship from becoming a contest over who understands the child better.

The question is not who wins the story.

The question is what new evidence helps the child.

62. The Probe Can Protect Family Life

A child struggles with homework every night.

The family assumes weak Mathematics.

A short supervised probe shows that the child solves the mathematics well when the task is started early and instructions are clear.

The evening problem may therefore belong partly to timing, fatigue or task initiation.

The Learning Dispatcher can now route the job differently.

Perhaps home needs a simpler routine rather than more teaching.

A good probe can prevent family conflict from being fed by the wrong academic explanation.

63. The Probe Should End in a Decision

After the answer, something should change.

Teach.

Practise.

Retest.

Route elsewhere.

Gather one more discriminating data point.

Or decide the original error was noise and leave it alone.

A Diagnostic Probe is not successful because the question was clever.

It is successful because the answer improves the next move.

64. The Probe Should Sometimes Produce “Still Uncertain”

One probe does not always settle the matter.

Maya passes the easy conceptual percentage question.

That weakens Story A.

It does not prove Story B completely.

Perhaps she succeeds only because the numbers are simple.

The next probe can restore multi-step structure while keeping the reference relationship explicit.

Diagnosis is often iterative.

The correct intellectual posture is:

Update confidence in the explanation; do not pretend one answer gives certainty the evidence does not contain.

65. The Probe and the Translation Layer Form a Pair

The Translation Layer says:

“Here is the precise behaviour we observed.”

The Diagnostic Probe says:

“Here are two plausible mechanisms. What fresh task separates them?”

Together they prevent a broad school signal from turning immediately into a broad tuition programme.

Signal becomes description.

Description becomes competing explanations.

Probe chooses evidence.

Evidence narrows the next job.

66. The Probe and the Priority Queue Form Another Pair

When the probe identifies a likely mechanism, The Priority Queue asks whether that mechanism deserves immediate work.

A diagnosis can be correct and still be low priority.

A rare error may be well understood but not worth twenty minutes.

A recurring upstream error may deserve the front of the queue.

Diagnosis tells us what.

Priority tells us when.

67. The Probe and the Learning Simulator

The Learning Simulator can create diagnostic conditions deliberately.

Suspect time pressure?

Simulate time.

Suspect cue dependency?

Remove cues.

Suspect representation?

Switch forms while preserving the concept.

The probe supplies the comparison logic.

The simulator supplies the controlled environment.

68. The Probe and the Flight Recorder

One probe is stronger when history exists.

The Flight Recorder tells the tutor whether the same condition produced the same failure last month.

That changes the competing explanations.

A sign error occurring once may be noise.

The same error occurring only when negatives and brackets interact becomes a stronger hypothesis.

The recorder supplies longitudinal clues.

The probe supplies targeted new evidence.

69. The Probe and the Intervention Threshold

A probe only works if the student gets to answer the question the probe was designed to ask.

If the tutor intervenes too early, the response stops being diagnostic.

During a probe, the Intervention Threshold may deliberately move later.

Let Maya commit.

Let Hana explain.

Let Jia Jun choose.

Then help.

The tutor should not answer the diagnostic question on the learner’s behalf.

70. The Probe and the Transfer Gate

The Transfer Gate can generate failures that need probing.

A skill fails under mixed conditions.

Why?

Recognition?

Retrieval?

Time?

Representation?

One diagnostic probe can identify the dimension that broke.

Then the next transfer test can target that boundary.

71. The Probe and the Exit Ramp

The Exit Ramp asks when support can shrink.

A Diagnostic Probe can prevent support from continuing for the wrong reason.

The student gets one bad mark.

Family thinks tuition must continue unchanged.

A probe shows the old target skill is stable; the error came from a one-off time decision.

Do not recreate the old programme.

Address the actual signal.

72. The Parent Can Ask a Diagnostic Question Without Becoming the Tutor

At home, parents can use one gentle question:

“Show me where it first stopped making sense.”

This is different from solving the problem.

The child may point to the instruction.

The first line.

The vocabulary.

The calculation.

Or say:

“I know how. I’m just tired.”

The parent does not need to diagnose completely.

They can preserve useful evidence and route it onward.

73. Do Not Turn Home Into a Diagnostic Laboratory

Every educational tool can be overused.

A parent reads this article and begins probing every error.

“Was it retrieval or representation?”

“Was that conceptual or procedural?”

Dinner becomes an assessment centre.

No.

Home should remain home.

The Diagnostic Probe is primarily a teaching instrument.

Parents need only enough clarity to avoid broad labels and preserve useful observations.

That keeps the Family Life Education boundary intact.

74. Teach the Student to Probe Their Own Errors

The adult should not own diagnosis forever.

An older learner can ask:

Did I not know it?

Or did I not recognise it?

Did I choose the wrong method?

Or choose correctly and execute badly?

Was the concept weak?

Or did time break it?

Can I test that distinction with one fresh example?

A student who can design a personal diagnostic probe is becoming a powerful independent learner.

75. The Student’s Four-Step Probe

Step 1 — Describe the error.
What exactly happened?

Step 2 — Name two possible causes.
Do not settle too quickly.

Step 3 — Change one thing.
Simplify language, remove time, provide the method, switch representation or ask for oral explanation.

Step 4 — Update the next action.
Teach, practise, retest or ask for help.

This is a compact form of scientific thinking applied to learning.

76. The Tutor’s Probe Checklist

• What exactly did the learner do?
• What are the two strongest plausible explanations?
• Would those explanations predict different behaviour under any simple condition?
• What is the smallest task that creates that condition?
• Can I remove irrelevant difficulty?
• Can I change only one important variable?
• Will the answer change my teaching decision?
• Am I accidentally cueing the answer?
• Do I need the learner to explain reasoning, not only select an option?
• Should all three students commit before discussion?
• What result would support each hypothesis?
• What result would leave the issue uncertain?
• What will I do immediately after the probe?

This is enough structure to keep diagnostic questioning disciplined without turning it into bureaucracy.

77. The Long Return: Adults Diagnose Before They Repair

Years later, Maya is not solving a percentage worksheet.

Something at work has failed.

A report is late.

Possible cause:

the process itself is too slow.

Alternative:

one approval step is creating the delay.

She does not rebuild the whole workflow immediately.

She checks where the delay occurs.

Another day, software fails.

Is the file corrupt?

Or does the programme fail on every file?

Open another file.

That is a diagnostic probe.

Competent adults do this constantly.

They change one thing, observe what happens and narrow the cause before deploying an expensive solution.

Tuition can teach that habit years before anyone calls it troubleshooting.

78. Back to Maya’s Percentage Question

The tutor now knows that Maya understands the reference idea in a simple form.

So he gives one more probe.

This time the problem is multi-step.

Before calculating, Maya must circle the reference amount.

She circles it correctly.

Then calculates correctly.

The tutor changes the wording.

She hesitates.

Circles the wrong quantity.

Now the boundary is visible.

Maya understands the conceptual relationship.

The difficulty appears when comparative wording obscures the reference frame.

The tutor does not reteach the entire percentage chapter.

He trains one decision:

Before computing a percentage change, identify what the change is being measured from.

Then fresh wording.

Then mixed questions.

Then delay.

Then school return.

The probe did not make the lesson more complicated.

It made the repair smaller.

That is the point.

---

The Diagnostic Probe in One Page

1. A wrong answer usually has more than one possible cause.
Do not teach the first plausible explanation automatically.

2. Describe the behaviour first.
Use the Translation Layer before forming explanations.

3. State two competing explanations.
A probe needs alternatives to discriminate.

4. Ask what each explanation predicts.
Find a condition under which the predictions diverge.

5. Change one important variable where possible.
Keep everything else simple.

6. Use simplification.
Reduce language, arithmetic, representation or time demands to see what remains.

7. Separate selection from execution.
Give the method or remove the calculation depending on what you need to test.

8. Change mode.
Oral versus written, graph versus prose, equation versus word problem.

9. Use contrast and counterexample.
Near-neighbours often reveal the rule more clearly than isolated questions.

10. Make distractors meaningful.
Wrong options should represent plausible misconceptions.

11. Protect the learner’s response.
Do not cue the answer before the diagnostic decision has been made.

12. Keep the probe cheap.
Use the smallest task that gives enough information.

13. Probe only when the distinction changes action.
Diagnostic curiosity is not enough.

14. Accept uncertainty.
One probe may change confidence without settling the explanation completely.

15. End with a better next move.
The value of the probe is the teaching decision it improves.

Where This Fits in eduKatePunggol

• How Tuition Works at eduKatePunggol — the broader tuition relationship.
• How Tuition Works | The Weak Link — identify the first limiting mechanism after enough evidence exists.
• How Tuition Works | The Learning Simulator — create controlled conditions for rehearsal and diagnostic comparison.
• How Tuition Works | The Sparring Partner — use resistance and counterexamples to strengthen and sometimes expose thinking.
• How Tuition Works | The Flight Recorder — preserve longitudinal evidence so one probe can be interpreted against history.
• How Tuition Works | The Learning Dispatcher — decide where the resulting learning job belongs.
• How Tuition Works | The Exit Ramp — reduce support as the learner becomes independent.
• How Tuition Works | The Priority Queue — decide which known problem deserves attention first.
• How Tuition Works | The Intervention Threshold — decide when to enter a live attempt.
• How Tuition Works | The Translation Layer — turn school and home signals into precise observable learning descriptions.
• How Tuition Works | The Transfer Gate — test whether capability survives changed conditions.
• Returned Paper Review — use authentic marked work to generate high-value diagnostic questions.

Batch 03 now has two distinct mechanisms:

• The Transfer Gate — how far does the learning travel?
• The Diagnostic Probe — when it fails, which cause best explains the boundary?

The sequence is becoming more precise:

Observe → translate → form competing explanations → probe → identify the weak mechanism → prioritise → teach → retest → transfer.

---

Research Foundations

The Diagnostic Probe is an explanatory tuition model. It draws on formative assessment, diagnostic assessment, hinge-question design and the broader principle that assessment should be selected according to the instructional decision it needs to inform.

• Education Endowment Foundation — Diagnostic Assessment Tool. EEF emphasises knowing why an assessment is being used, what information it is designed to produce and how that information will affect subsequent decisions. Its hinge-question guidance notes that carefully designed wrong answers can reflect specific misconceptions.
• Education Endowment Foundation — How Checking for Understanding Can Guide Teaching (2026). The guidance describes diagnostic hinge questions whose response options represent different misconceptions or levels of understanding so teachers can decide whether to pause, adapt support or move on.
• Dylan Wiliam — Designing Great Hinge Questions, Educational Leadership (2015). Wiliam describes questions designed to reveal understanding rapidly enough to determine the next instructional move and stresses avoiding questions where a learner can be correct for the wrong reason.
• Dylan Wiliam — The Right Questions, the Right Way, Educational Leadership (2014). The article argues for planning questions around common misunderstandings and collecting responses in ways that reveal more than the thinking of the most confident learner.
• Institute of Education Sciences — Understanding the Educational Assessment Landscape. IES recommends starting with the decision to be made, matching the assessment to that purpose and recognising what the chosen evidence cannot tell you.
• Institute of Education Sciences — Formative Assessment and Elementary School Student Academic Achievement. The review defines formative assessment as gathering, interpreting and using evidence about learning to support timely instructional adjustments.
• Institute of Education Sciences — Developing and Using Diagnostic Items in Mathematics and Science. The research programme focused on creating questions that identify student misconceptions and help teachers use those responses to improve instruction.

The practical conclusion is simple.

When a student gets something wrong, resist the pleasure of having an immediate explanation.

Ask whether another explanation still fits.

If it does, do not teach both.

Design one small question that makes them disagree.

Then listen to what the learner’s answer tells you.