BTT Mathematical Lab | Observe, Probe, Repair, Validate and Transfer

Quick Read: the Mathematical Lab loop

Evidence → Observation → Competing explanations → Probe → Minimum useful intervention → Reattempt → Validation → Transfer → Retention → Independence.

A failed validation is not wasted work. It is new evidence. If the student improves only on the taught example but fails a changed version, the original explanation was incomplete or the repair was too narrow. The Lab reopens the question instead of forcing a success story.

Why a Mathematical Lab is needed

A final answer hides the path that produced it. Two students can write the same wrong answer for completely different reasons. One may not understand the concept. Another may understand the concept but lose control during algebra. A third may read the condition incorrectly. A fourth may retrieve the wrong method because the question looks similar to something practised before. A fifth may know what to do but make an execution error under time pressure.

If all five students receive the same correction sheet, the teaching is efficient only in appearance. The Mathematics Lab exists to prevent that collapse of different causes into one vague label such as “careless”, “weak”, “slow” or “does not understand”. Those labels describe an outcome. They do not identify the mathematical mechanism.

The Lab asks a narrower and more useful question: What is the earliest important point at which this attempt stops being reliable? Once that point is located, the tutor can decide whether the next move is explanation, representation, a prerequisite repair, deliberate practice, a transfer task, examination rehearsal or simply less help.

Observation before interpretation

The first discipline of the Lab is to describe what happened before deciding why it happened. “The student is careless” is an interpretation. “The student copied +7 as −7 on the second line and did not check the substitution at the end” is an observation. “The student does not understand fractions” is an interpretation. “The student can compare unit fractions but reverses the comparison when the numerators differ” is an observation.

• Read the question: What does the learner underline, ignore or misread?
• Watch the start: Does the learner choose a representation or method immediately, hesitate, or begin with an unrelated procedure?
• Inspect the first wrong line: The earliest departure from valid Mathematics is often more informative than the final answer.
• Notice recovery: Does the learner detect that something is wrong? Can they recover after a prompt?
• Check explanation: Can the learner say what a symbol, operation, graph, ratio or derivative means?
• Reduce support: What happens when the example is removed, the numbers change or the wording is unfamiliar?

Good observation is deliberately modest. It records evidence without claiming more than the evidence can support.

One visible error can have several causes

Visible problem	Plausible mathematical causes	Useful next probe
Wrong answer to an algebra equation	Operation meaning, negative-number control, inverse-operation misconception, transcription error, weak checking	Use a simpler equation, then a structurally similar equation with different surface numbers
Cannot solve a word problem	Reading target, quantity relationships, representation, prerequisite arithmetic, method selection	Ask the learner to represent the situation without calculating
Forgets a method	Weak understanding, insufficient retrieval practice, cue dependence, poor discrimination between similar methods	Ask for the method in a changed context without a chapter label
Slow in A-Math	Fragile algebra, method-selection uncertainty, excessive checking, poor symbolic fluency, unfamiliar mixed-topic structure	Separate algebraic execution speed from method-selection time
Loses marks under examination conditions	Knowledge gap, retrieval failure, pacing, paper navigation, notation, checking cost, pressure-related execution breakdown	Compare untimed and timed attempts on matched Mathematics

The purpose of a probe is not to make the student fail. It is to distinguish between explanations that would require different teaching.

Form competing explanations, not one favourite story

A common teaching error is to decide too early. If a tutor believes the problem is algebra, every new mistake can be interpreted as further proof of weak algebra. If a parent believes the child is rushing, every lost mark can become “carelessness”. The Lab resists that confirmation loop.

Instead, keep a small set of plausible explanations open. For example, a student who cannot complete a trigonometric equation may have weak trigonometric identities, weak algebraic manipulation, poor recognition of the target form, or difficulty managing several transformations in working memory. A good probe changes one part of the task while holding the others as stable as possible.

The best next question is not always the hardest question. It is the question that gives the most useful information about what to do next.

How to design a discriminating Mathematics probe

1. State the competing explanations. What two or three causes would lead to different teaching?
2. Change one important demand. Remove difficult arithmetic, simplify the diagram, name the method, remove the chapter label, or change the representation.
3. Preserve the mathematical relationship. The probe should still test the same underlying structure where possible.
4. Ask for working and explanation. A final answer alone often cannot distinguish causes.
5. Use the result cautiously. One probe changes confidence; it rarely proves a complete learner model.

Suppose a student fails a percentage reverse problem. Before reteaching the entire percentage chapter, ask a simpler question that separates percentage-as-operator from algebraic reversal. If the learner can find 20% of a quantity but cannot reconstruct the original from 80%, the repair should target inverse multiplicative structure rather than basic percentage calculation.

Minimum justified help

Once the likely weak link is identified, the next question is how much help to give. More help is not automatically better. Help should be sufficient to restart productive mathematical activity while preserving as much learner thinking as possible.

• Wait: when the learner is still productively thinking.
• Ask a question: when attention needs to move to a relevant condition or relationship.
• Change the representation: when the idea may be present but inaccessible in the current form.
• Give a cue: when retrieval is the main problem.
• Give a partial hint: when the learner needs a small restart but can still carry the method.
• Model one step: when a transformation or notation convention needs to be made visible.
• Teach directly: when the required idea is genuinely missing or seriously misconceived.

After help, the tutor should remove it. A successful supported solution is not the end of the experiment. The important evidence comes from the next independent attempt.

Repair the relationship, not only the question

Weak teaching repairs an item. Strong teaching repairs the relationship that made the item difficult. If a learner incorrectly expands 3(x − 4), the important question is whether the error came from distribution, negative-number control or inattentive copying. If distribution is weak, the repair should connect multiplication to the entire grouped quantity. If negative numbers are weak, the repair belongs earlier. If the learner knows both but copied the sign incorrectly, the intervention should focus on execution and verification rather than reteaching distributivity.

The Mathematics Knowledge Warehouse is useful here because it organises ideas by mathematical object rather than school chapter. The Lab can move backward to a prerequisite without pretending that the student must repeat an entire previous year.

Reattempt: the first test of the repair

The first reattempt should be close enough to the original problem that the repaired relationship is still relevant, but not so identical that the learner can simply copy a remembered sequence. Change the numbers, order, representation or wording while preserving the important structure.

If the learner succeeds only when the tutor’s example remains visible, the repair is not yet independent. If the learner succeeds after one prompt but cannot start alone, the issue may have shifted from understanding to retrieval or recognition. The Lab treats these as different states.

Validation: did the repair actually work?

A repair earns confidence through evidence. The learner should be able to explain the corrected idea, use it in a changed question, reconnect it to the current topic and complete enough of the work without the tutor carrying the decision-making.

• Concept validation: Can the learner explain the relationship?
• Execution validation: Can the learner carry the method accurately?
• Recognition validation: Can the learner identify when the idea is relevant?
• Transfer validation: Can the learner use the idea when the surface changes?
• Independence validation: Does performance survive reduced prompting?

If validation fails, reopen the explanation. Do not simply prescribe more repetitions of the same task.

Transfer: can the Mathematics survive a changed surface?

School practice often groups similar questions together. That is useful during early learning, but it can create an illusion of mastery because the chapter title and neighbouring examples tell the learner what method to use. Transfer removes those supports.

A transfer task changes something meaningful: the wording, representation, order of information, diagram orientation, numbers, context or combination of topics. The underlying mathematical relationship remains recognisable to someone who understands it, but the surface no longer announces the method.

Transfer is where a repaired idea becomes usable Mathematics rather than a repaired worksheet pattern.

Retention: is the repair still available later?

Immediate success can be misleading. The learner may still be using the tutor’s explanation from working memory. A later return asks whether the Mathematics has become available again after time, other topics and competing information have intervened.

Retention does not require repeating every diagnostic exercise indefinitely. Important ideas can be sampled later through spaced retrieval, mixed practice and natural use inside later Mathematics. The strength of the return check should match the importance of the dependency.

Release: when the Lab should stop

The purpose of the Lab is not to keep a learner permanently under observation. It should release the learner when the relevant relationship is stable enough for normal learning to continue. The student returns to the school-stage route, Knowledge Warehouse object, practice system, examination preparation or independent work that now makes sense.

A good diagnostic process should reduce dependence on diagnosis.

Worked Lab Example 1: the same algebra mistake, two different repairs

Consider two students solving an equation that requires expanding brackets and collecting like terms. Both write the wrong sign after expansion.

Student A consistently treats subtraction outside a bracket as if it changes only the first term. When asked to expand −2(x − 3) using a rectangle/area representation and then by repeated addition, the same misconception remains. The likely weak link is distributive structure with signed multiplication. The repair belongs in the mathematical relationship.

Student B expands three similar expressions correctly, explains distribution accurately and catches the error immediately when asked to substitute a value. The original sign loss is better treated as execution instability and weak verification. Reteaching distribution would add volume without addressing the real problem.

The visible error was the same. The Lab prevents the teaching response from being the same.

Worked Lab Example 2: a Primary word problem that is not really an arithmetic problem

A learner can add, subtract, multiply and divide accurately in isolated exercises but repeatedly chooses the wrong operation in multi-step word problems. Instead of assigning more mixed worksheets immediately, the tutor removes calculation and asks the learner to draw or describe the quantities and relationships.

If the learner cannot identify what is being compared or what remains unchanged, the weak link may be representation and relationship modelling. If the learner represents the situation correctly but later calculates inaccurately, the problem is different. The Lab separates the model from the arithmetic before prescribing practice.

Worked Lab Example 3: A-Math differentiation that collapses under algebra

A Secondary student can explain gradient and can differentiate simple powers, but fails on a more complex differentiation question. It is tempting to conclude that differentiation is weak. The Lab isolates the layers.

First, give a differentiation task with very light algebra. If that succeeds, increase algebraic complexity without changing the calculus idea. If performance collapses only when factorisation, fractions or symbolic simplification become demanding, the calculus concept may be carrying an algebra bottleneck. The repair should increase symbolic bandwidth rather than repeat introductory differentiation notes.

This is why the Algebra → Functions → Calculus dependency corridor matters. Later Mathematics can expose an earlier constraint without being the original cause of it.

Worked Lab Example 4: examination underperformance

A student performs well in lessons and topical practice but loses large numbers of marks in full papers. The Lab compares matched work under different conditions. If untimed mixed questions are also weak, the issue is not primarily examination pressure; topic integration or recognition may still be unstable. If untimed mixed work is strong but timed performance deteriorates, the tutor can examine pacing, method-selection time, checking cost and recovery after difficult questions.

That investigation then hands over to Mathematics Examination Craft, where paper navigation, timing, retrieval and checking are treated as an execution system rather than as generic advice to “practise more papers”.

The Mathematical Lab in a three-student class

A three-student tutorial creates useful comparison, but the Lab must keep the learners separate. One student’s error should not automatically become the diagnosis for the group. The tutor can use the same mathematical prompt while observing different starting choices, representations and recovery patterns.

• One learner may need a prerequisite repair.
• One may need less prompting so recognition develops.
• One may already understand the topic and need transfer or examination pressure.

The group creates shared mathematical discussion; the evidence still belongs to each learner. That distinction is central to the three-student model.

What parents can bring to the Lab

Parents do not need to arrive with a diagnosis. Useful evidence can be much simpler: a recent marked paper, school feedback, examples of corrections, a description of where homework stalls, or a pattern such as “understands during tuition but cannot start alone two days later”.

The Lab converts that information into mathematical questions. It does not ask parents to decide whether the problem is memory, confidence, algebra or examination technique before the work has been examined.

What students can bring to the Lab

The most useful student contribution is often the unfinished or imperfect attempt. Do not erase every wrong line before asking for help. The route into an error can show the tutor what the final answer cannot.

• Bring the question and your actual working.
• Mark the line where you stopped being sure.
• Say what you were trying to do, even if the explanation is incomplete.
• If you changed your mind, show the earlier attempt rather than hiding it.
• After help, attempt a changed question without copying the previous solution.

Teacher judgement, AI and evidence

AI can be useful for generating alternative explanations, proposing discriminating questions, comparing solution paths and surfacing overlooked mathematical dependencies. It should not be treated as certainty. A teacher has contextual information, live observation and responsibility for the educational decision; a model has pattern-recognition strengths and important limits.

When teacher and AI interpretations disagree, the Lab does not solve the disagreement by choosing the more confident voice. It asks what additional mathematical evidence would distinguish the explanations. The Teacher and AI Agreement Testing guide explains this principle in more detail.

What the Lab deliberately does not claim

• It does not infer a medical, psychological or neurodevelopmental condition from Mathematics work.
• It does not treat one wrong answer as proof of a stable learner trait.
• It does not assume every weak mark requires tuition.
• It does not replace school teachers, official syllabuses or examination authorities.
• It does not promise that one intervention will work for every learner.
• It does not turn a temporary support strategy into permanent dependence.

How the Mathematical Lab connects to the BTT Mathematics system

Question	Best route
Where does the learner’s route first become unreliable?	Mathematics Diagnosis
What mathematical object or prerequisite is involved?	Mathematics Knowledge Warehouse
How much help should be given?	How Mathematics Help Works
How should an uncertain explanation be tested?	Mathematical Lab
How should knowledge be converted into examination performance?	Mathematics Examination Craft
Where should the learner continue after repair?	Mathematics Learning Library

A compact Lab protocol for Mathematics tutors

1. Begin with the learner’s actual attempt.
2. Record the earliest reliable observation before naming a cause.
3. Keep more than one plausible explanation open when the evidence is ambiguous.
4. Choose the smallest probe that would change the teaching decision.
5. Use the least help that restarts productive Mathematics.
6. Require a fresh attempt.
7. Change the surface and test transfer.
8. Return later when retention matters.
9. Reduce support when the learner can carry the relationship independently.
10. If the evidence contradicts the original explanation, reopen the investigation.

Frequently asked questions

Is the Mathematical Lab a diagnostic test?

No. It is a method for investigating Mathematics learning through real attempts, carefully chosen probes and validation. A formal test can be one source of evidence, but the Lab is not defined by a single paper or score.

Does every student need the Lab?

No. If the problem is already clear and the appropriate teaching response is obvious, there is no reason to create an investigation. The Lab is most useful when the visible problem has several plausible causes, previous repairs have not held, or performance changes sharply across contexts.

How many questions does a diagnostic probe need?

As few as can distinguish the relevant explanations with reasonable confidence. The objective is information, not testing volume.

What happens if the learner improves immediately?

Immediate improvement is encouraging but incomplete evidence. Use a fresh example, reduce support, change the surface and return later if the dependency is important.

What happens if the repair fails?

Reopen the explanation. A failed repair may show that the original cause was wrong, that two constraints interact, or that the intervention did not address the relationship strongly enough.

Where does the learner go after the Lab?

Back into normal Mathematics: the relevant school-stage route, Knowledge Warehouse object, practice sequence, examination preparation or independent work. The Lab is a temporary investigation layer, not a permanent destination.

Mathematics system route: Mathematics Hub · Diagnosis · Knowledge Warehouse · Mathematics Help · Examination Craft · Learning Library.