The Tutor Handbook Vol No.0031 | The New Baseline — How a Tutor Rebuilds the Learner Model After a Weak Link Has Needed Repair Twice

The Tutor Handbook · Volume 0031 · Series ID THB-0031

The same weak link has now needed two genuine repairs.

That fact matters.

It does not mean the learner is permanently weak. It does not mean the first repair was fake. It does not mean the second repair will fail. It does not mean the tutor should turn every future lesson into surveillance of one old problem.

But it does mean the old learner model is no longer good enough.

A tutor who once believed, “This capability is repaired and can now be treated as ordinary,” has received new evidence. The capability may still be strong. It may still transfer. It may still survive long periods. Yet its history now tells us something more precise: under some combination of time, load, neglect, transition, representation, stress, curriculum growth or reduced practice, this part of the learner’s system has shown a tendency to become limiting again.

The correct response is not fear.

It is a better map.

This is Volume 0031 of The Tutor Handbook, eduKate Sengkang’s long-form series on the practical decisions inside tutoring.

Volume 0030, The Second Repair, owned the design of a new intervention after a previously repaired weak link had reopened. It asked what survived, what changed, what should be reused, what should be modified and what should be replaced.

This volume begins after that second repair has produced enough evidence to update the tutor’s view of the learner.

The new baseline is the tutor’s current, evidence-based working model of what the learner can reliably do, under which conditions, with how much support, at what cost, and with which known fragilities after the learner’s history has changed.

That is narrower than a general baseline article. Baseline Before Intervention owns the problem of establishing what a learner could do before an intervention begins. Bolt 31 — The Difference Between a Peak and a Baseline owns the measurement distinction between exceptional performance and ordinary performance level. This handbook volume owns a different tutor decision: once a learner has accumulated a meaningful repair history, how should the tutor rebuild the learner model so future teaching starts from reality rather than from an expired picture?

The broader mechanisms retain their canonical owners. How Learning Diagnosis Works owns diagnosis. Finding the First Weak Link in Learning owns the wider weak-link model. How Learning Calibration Works owns calibration. How Memory Works in Learning owns the wider memory mechanisms. How Learning Transfer Works owns transfer. How Studying Works and How Self-Directed Studying Works own the learner’s deliberate study activity. eduKate Sengkang Education Runtime owns the wider educational system around the learner. How to Improve Anything owns the general improvement loop. The canonical Tutor Classification Model by eduKateSG owns the Class 0–6 tutor ladder.

This handbook chapter has one practical job:

After a weak link has required repair twice, what should the tutor now believe about the learner—and what should that updated belief change?

Quick Read

A new baseline is a working learner model, not a new label.
Two repairs are evidence that the old model needs updating, not proof of permanent weakness.
Record capability and fragility separately: what the learner can do, and the conditions under which that capability tends to weaken.
Do not use one mark as a baseline. Use several comparable pieces of evidence across time, support levels and task conditions.
Separate floor, ordinary working range and ceiling. A peak is not the baseline, and a bad day is not automatically the baseline either.
Track support demand. Correct work that needs repeated prompting is a different learner state from correct independent work.
Track recovery cost. A learner who can self-correct after one cue is different from one who needs a full reteach.
The new baseline should alter maintenance only where evidence justifies it.
Known fragility should become a watchpoint, not the identity of the student.
Class 3 Diagnostic Tutor work updates the mechanism model; Class 4 Route Designer work changes the future route; Class 5 Performance Coach work tests reliability under load; Class 6 Learning Architect work integrates the updated model across studying, transitions and independence.
Learning, studying, training, education and improvement remain different jobs and should be updated separately.
The model stays provisional. Future evidence can strengthen, narrow or retire the fragility hypothesis.

1. The Old Baseline Has Expired

A baseline is useful only while it predicts the learner well enough to guide decisions.

Before the first repair, the tutor may have believed:

This student cannot yet do X reliably.

After the first repair, the model changed:

This student can now do X independently across ordinary tasks.

After a long period of success, the tutor may reasonably have reduced monitoring. Then the capability reopened. A second repair was needed. The learner recovered again.

The tutor cannot simply return to the post-first-repair sentence as though nothing happened. The history now contains evidence that reliability is conditional in a way the first model did not capture.

The old baseline is not morally wrong. It is informationally incomplete.

2. A Baseline Is Not a Verdict

The word baseline can sound fixed.

It should not.

A tutor baseline is a working reference point. It helps decide what the learner can normally carry, what needs support, what can be stretched and what deserves observation. It should change when sufficiently good evidence changes.

That means the learner is never sentenced to the baseline.

The baseline is a map of the current terrain, not a fence around the learner’s future.

3. Marks Are Evidence, Not the Whole Model

A tutor who updates the baseline from marks alone will often miss the mechanism that matters.

Two students can both score 72% and have different learner states.

One worked independently, selected methods well, checked errors and lost marks on two unfamiliar extensions.
One needed heavy prompting during preparation, relied on topic labels and collapsed whenever a familiar surface changed.

The score is the same. The model should not be.

Marks matter because they are consequences produced under particular conditions. But a useful tutor baseline also records the route that produced them.

4. Build a Conditional Baseline

The strongest learner models are conditional.

Instead of:

She is good at algebra.

write something like:

She is reliable on algebraic manipulation in isolated and mixed untimed work, usually self-corrects sign errors, but accuracy drops in long calculus chains when working-space compression increases.

The second sentence is less elegant. It is much more useful.

It tells the next tutor, the learner and the future route designer what can be trusted and where reality has previously exposed a boundary.

5. Capability and Fragility Must Be Recorded Separately

A recurring weak link creates a dangerous temptation: let the fragility erase the capability.

Do not write:

Weak at fractions.

if the learner can explain fraction meaning, generate equivalents, compare values and use fractions correctly across most ordinary work.

Write both sides.

Capability	Known fragility
Understands equivalent fractions and uses them independently in ordinary problems	Selection slows when fractions are embedded among ratio, percentage and decimal representations under time pressure
Uses inference boundaries accurately on short and medium texts	Evidence checking tends to disappear late in long papers
Solves equations accurately	Sign control becomes vulnerable when several transformations are compressed mentally

This prevents a known vulnerability from swallowing the learner’s actual strength.

6. The New Baseline Needs Three Levels: Floor, Working Range and Ceiling

A single number cannot describe a changing learner very well.

Floor: what remains available even on a weaker day or under mild disruption.
Working range: what the learner normally produces under ordinary realistic conditions.
Ceiling: what appears when preparation, attention, task fit and execution all align well.

A second repair is especially informative because it can show whether the floor moved, whether the ordinary range became more stable, and whether the ceiling now rests on a stronger structure.

The goal is not to turn every lesson into measurement. It is to stop one exceptional success or one exceptional failure from becoming the entire learner story.

7. The New Baseline Should Include Recovery Cost

Two learners can both experience a recurrence and still have very different systems.

Learner A needs one cue and immediately reconstructs the route.

Learner B needs several lessons of concept rebuilding, guided examples, practice and transfer work.

Both had a weak link reopen. Their recovery cost is not the same.

How hard the system is to restore is itself part of the learner model.

Recovery cost helps the tutor decide how early to intervene next time, how much maintenance is justified and how dangerous it would be to ignore the first signs of drift.

8. Support Demand Is Part of the Baseline

Correctness without support information is incomplete.

A learner who can solve five questions after the tutor says, “Which formula?” is not in the same state as a learner who independently identifies the structure, chooses the formula and checks the answer.

The new baseline should therefore distinguish:

full modelling;
partial modelling;
directional prompts;
simple signals that checking is needed;
independent execution;
independent detection and correction of errors.

One of the most important signs of genuine second repair is not simply that the answer is right again, but that the tutor can become less visible without the capability disappearing.

9. Conditions Belong Inside the Learner Model

Capability changes with conditions.

fresh versus fatigued;
untimed versus timed;
single-topic versus mixed;
familiar surface versus changed surface;
short task versus long paper;
immediate practice versus delayed return;
with notes versus without notes;
after a model versus after no model;
quiet room versus realistic school pressure;
one representation versus another.

The point is not to excuse failure by inventing conditions after the fact. The point is to identify repeatable patterns.

If the same capability repeatedly fails only under long mixed timed work, that condition belongs in the model because it changes what the tutor should train.

10. Use an Evidence Hierarchy

Not every successful performance deserves equal weight.

Evidence	What it can tell the tutor
Immediate supported success	The learner can participate in the route with help
Fresh independent item	The learner can generate the route without the model visible
Changed surface	The learner can carry the structure beyond one familiar appearance
Mixed set	The learner can choose among competing routes
Delayed return	The capability remains accessible after recency fades
Current school task	The repaired capability reconnects to the educational route
Realistic timed performance	The learner can express the capability under relevant load
Repeated independent evidence	The result is becoming ordinary enough to inform the baseline

The new baseline should be built from the strongest available evidence, not from the most convenient evidence.

11. One Good Lesson Cannot Carry the Whole Update

The second repair may finish with a beautiful lesson.

The learner explains the rule, solves the fresh problem, handles the changed surface and leaves feeling strong.

That is encouraging.

It is not yet enough to rebuild the long-term baseline.

The tutor needs later evidence because immediate performance can be inflated by recent explanation, active cues, task similarity and focused attention.

The new model should emerge across several returns, not from one triumphant ending.

12. One Bad Day Cannot Carry the Whole Update Either

The opposite error is equally common.

A learner performs badly once after the second repair and adults immediately conclude that nothing has changed.

Before lowering the baseline, ask:

Was the task comparable?
Was the learner ill, exhausted or unusually distracted?
Did the failure appear in the same mechanism?
Was the task much harder?
Did a new prerequisite enter?
Could the learner recover independently once the error was noticed?

A baseline should be stable enough to resist ordinary noise and flexible enough to respond to real change.

13. Two Repairs Change the Forecast, Not the Identity

A recurring weak link should affect the tutor’s forecast.

Perhaps this capability deserves a slightly earlier check during major transitions. Perhaps maintenance should contain one mixed item every few weeks. Perhaps full-paper load should be introduced with more attention to the known failure mode.

That is a forecast update.

It is not the same as telling the learner:

You are the student who always fails here.

The first statement changes monitoring. The second changes identity. Good tutoring needs the first and should be extremely cautious with the second.

14. The Model Should Explain Why the Second Repair Was Needed

Do not record only that a second repair happened.

Record the best current explanation.

Low-use retrieval decay?
New curriculum increased the processing load?
Transfer range remained too narrow?
Method selection failed when labels disappeared?
Exam timing exposed an execution weakness?
Independent studying stopped revisiting the skill?
A transition changed representations or notation?
Support had faded before the learner could self-correct?

The mechanism matters because different causes justify different future safeguards.

15. The First Repair and Second Repair Should Be Compared

Question	Why it matters
Was the second repair shorter?	Prior learning may have survived even though access weakened
Was less scaffolding needed?	The learner may now own more of the route
Did the same mechanism fail?	The fragility may be specific and repeatable
Did a different mechanism fail?	The visible symptom may hide a changing system
Did transfer fail sooner than concept knowledge?	Maintenance may need wider surfaces rather than more explanation
Did performance fail only under load?	Future work may belong to performance training rather than reteaching
Did the learner self-diagnose sooner?	Metacognitive control may have improved even when the academic weakness reopened

A second repair is not merely another intervention. It is a second experiment on the same changing system.

16. Update the Fragility Hypothesis Proportionately

Suppose a capability was stable for eighteen months, weakened during a curriculum transition, and returned quickly after a short targeted repair.

That history does not justify treating the capability as chronically unstable every week.

Now suppose the same capability has reopened twice within one term under ordinary work and each repair required extensive support.

That history deserves more weight.

The baseline should move by the weight of the evidence, not by the emotional intensity of the latest event.

17. Keep the Model Falsifiable

A useful learner model can be proved wrong.

For example:

Current hypothesis: inference discipline is reliable in ordinary reading but becomes fragile when long-passage load and time pressure combine.

Now the tutor can test that claim. If the learner performs reliably across several long timed passages, the fragility hypothesis should weaken. If failures repeatedly appear in exactly that condition, the hypothesis gains weight.

Do not preserve a tutor theory simply because it once sounded intelligent.

The learner model must remain answerable to the learner.

18. The Baseline Should Include Self-Correction

A learner who never makes a mistake is not the only picture of strength.

Sometimes a stronger sign is what happens after the mistake.

Does the learner notice the inconsistency?
Can they locate the line where the reasoning changed?
Can they name the violated rule?
Can they repair without being told the full answer?
Can they prevent the error on a fresh item?

Two repairs may actually leave the learner more resilient than before if the learner now knows how to detect and restore the system independently.

19. The Baseline Should Include Initiation

Academic capability can be present while independent initiation remains weak.

A learner may know exactly how to review corrections, retrieve vocabulary or practise algebra, yet do none of it until an adult starts the process.

If the second recurrence happened because the learner’s study system stopped revisiting the skill, initiation belongs in the updated model.

That does not mean the tutor should become the permanent starter motor.

It means the next route needs to build a start mechanism the learner can own.

20. The Baseline Should Include Selection

Many learners can perform methods they fail to select.

A student can solve simultaneous equations, factorise, use trigonometry and manipulate ratios when the worksheet heading announces the topic. In a mixed paper, the same learner may choose badly.

If this selection problem has required repair twice, the new baseline should not say, “Methods mastered.”

It should distinguish execution from routing.

Knowing how to use a tool is not the same as recognising when the tool belongs.

21. The Baseline Should Include Transfer

A repaired skill that works only in the tutor’s familiar format has not earned a broad baseline claim.

If the weak link reopened when the school changed wording, diagram style, representation, genre or question structure, record the transfer boundary.

Then test whether the second repair widened it.

same rule, different wording;
same relationship, different numbers;
same concept, different representation;
same skill, different subject context;
same method, no topic label;
same principle, larger task.

The new baseline should describe where learning travels, not just where it sits comfortably.

22. The Baseline Should Include Performance Load

There is a difference between having knowledge and delivering it when the conditions become crowded.

Long papers introduce timing, task switching, fatigue, error recovery, decision pressure and the need to allocate attention economically.

If a learner has needed two repairs because available knowledge repeatedly collapses under load, the new baseline should state that clearly.

Then future work can target the real boundary rather than pretending the student has forgotten the whole subject.

23. Learning, Studying, Training, Education and Improvement Must Be Updated Separately

The eduKate architecture keeps these words separate because each asks a different question.

Learning: What usable capability now exists? What can the learner retrieve, explain, choose, transfer and repair?

Studying: What deliberate learner activity keeps that capability accessible without tutor orchestration?

Training: Under which repeated conditions should the learner practise so the vulnerable boundary becomes more robust?

Education: How does the learner’s current school, syllabus, assessment and transition route create the next demands?

Improvement: What limiting mechanism should change next, what evidence would show change, and when should the system be reviewed again?

A new baseline can change one of these without changing all five.

24. A Learning Update Does Not Automatically Require More Studying

Suppose the second repair shows that conceptual knowledge is now strong, retrieval is fast and transfer is broad, but performance becomes less accurate only during the last twenty minutes of a long paper.

The answer is not necessarily “study the topic more”.

The learner may need performance training: pacing, working-space discipline, fatigue management and error-recovery routines.

The new baseline helps the tutor stop sending work to the wrong layer.

25. A Studying Update Should Change What Happens Between Lessons

If the recurring weak link is partly caused by low-use decay, the outside-lesson route matters.

But “more homework” is too blunt.

one retrieval item after a delay;
one changed-surface item at the end of another topic;
one correction reattempt without notes;
one short mixed set rather than a long same-type worksheet;
one learner-generated explanation of the fragile decision point.

The study plan should preserve the repaired capability with the lowest useful maintenance cost.

26. A Training Update Should Target the Boundary

Training is useful when repeated exposure is aimed at a precise future demand.

If method selection is fragile, train discrimination.

If transfer is fragile, vary the surface.

If retrieval is fragile, space the returns.

If performance is fragile, increase realistic load gradually.

If self-correction is fragile, let the learner find and repair errors before feedback takes over.

Training should make the known boundary less costly. It should not become endless repetition of what is already secure.

27. An Education Update Changes the Route, Not Just the Worksheet

The learner’s school path may explain why the old baseline expired.

A move from Primary to Secondary changes abstraction, notation, independence and workload. A move into Additional Mathematics increases algebraic density. A move into examination-year full papers increases timing and integration. A move into JC compresses the time available for repair while increasing conceptual and procedural load.

The updated learner model must sit inside that educational route.

The same learner can require a different baseline description when the world asking the questions has changed.

28. An Improvement Update Changes the Next Experiment

A good baseline is not merely descriptive.

It improves the next decision.

If the learner’s new baseline says that transfer is broad but timed selection remains fragile, the next experiment should test timed discrimination rather than conceptual reteaching.

If the learner’s baseline says that the capability is reliable and self-correcting across realistic conditions, the next experiment may be to reduce monitoring further.

The model earns its place by helping the tutor choose what to do—and what not to do.

29. The Tutor Classification Model Becomes More Important After Repeated Repair

The canonical tutor ladder remains: Class 0 Homework Helper, Class 1 Explainer, Class 2 Drill Builder, Class 3 Diagnostic Tutor, Class 4 Route Designer, Class 5 Performance Coach and Class 6 Learning Architect.

Class	New-baseline job
0 · Homework Helper	Supports routine work without hiding whether the fragile capability is becoming dependent again
1 · Explainer	Clarifies only genuinely unclear relationships; does not treat every recurrence as a concept problem
2 · Drill Builder	Stabilises retrieval, fluency or discrimination when repetition actually fits the evidence
3 · Diagnostic Tutor	Updates the mechanism model and distinguishes recurrence from a different new bottleneck
4 · Route Designer	Changes sequencing, maintenance and transition planning around the updated model
5 · Performance Coach	Tests whether the capability survives realistic timing, integration and pressure
6 · Learning Architect	Integrates the new baseline across learning, studying, independence, family communication and long-term development

30. Class 3: Rebuild the Mechanism Model

The Diagnostic Tutor asks whether the second recurrence was truly the same break.

The visible symptom may be familiar while the underlying cause has moved.

A student who once lacked vocabulary may now know the vocabulary but fail to control answer scope. A student who once lacked equation balance may now understand it but lose signs under compressed multi-step work. A student who once needed help starting homework may now start independently but avoid corrections.

The diagnostic update prevents old labels from becoming self-fulfilling explanations.

31. Class 4: Rebuild the Route

Once the model changes, the route may need to change.

Introduce mixed work earlier.
Keep one known-fragility item in spaced review.
Delay full-paper load until one prerequisite is stronger.
Bring full-paper load forward because the foundation is now clearly stable.
Move the learner from weekly maintenance to monthly sampling.
Reduce remedial time because recovery is now fast and independent.

Route design turns the learner model into a calendar of better decisions.

32. Class 5: Rebuild the Performance Forecast

A capability can be academically repaired and still be performance-fragile.

The Performance Coach asks:

Does accuracy survive time pressure?
Does method selection survive mixed papers?
Does the learner recover after one difficult item?
Does checking still happen late in the paper?
Does the repaired skill survive fatigue?
Can the learner allocate time without tutor reminders?

This creates a performance baseline rather than pretending classroom competence and examination competence are interchangeable.

33. Class 6: Rebuild the Whole Learner Model

The Learning Architect has the widest job.

The second recurrence may connect to more than one layer: academic structure, study behaviour, workload, confidence, transition, family support, school timing and the learner’s ability to self-monitor.

Class 6 work integrates those layers without turning the learner into a permanent repair project.

The whole model should become more precise while the learner becomes more free.

34. A New Baseline Should Shrink Tutor Dependence, Not Expand It

Once adults know a weakness has reopened twice, they may become reluctant to let go.

Every answer is checked. Every homework session is supervised. Every small error triggers intervention.

This can make the learner look safer while quietly reducing independence.

The new baseline should instead specify what the learner can now own.

self-check;
self-correction;
retrieval schedule;
error classification;
first response to drift;
when to ask for help.

A known fragility should create better self-management, not permanent adult occupation.

35. The Six eduKate Sengkang Learners: Six New Baselines

Alicia: retrieval has weakened twice after long gaps. Her new baseline records strong understanding and fast relearning, with low-use retrieval as the known fragility. Maintenance becomes brief spaced generation rather than repeated reteaching.
Beatrice: two recurrences appeared only when similar methods were mixed. Her baseline separates strong execution from fragile discrimination. Her route keeps contrast pairs alive across the term.
Ciara: inference discipline is reliable until passage length and timing combine. Her baseline includes the load boundary and her ability to recover by anchoring evidence.
Denise: content remains strong but late-paper accuracy has failed twice. Her updated model is primarily about performance stamina, pacing and checking—not conceptual weakness.
Emily: the academic skill survives, but independent review disappears during busy periods. Her baseline includes study initiation as the recurrent weak control point.
Faith: the concept itself genuinely weakened twice. Her baseline records slower recovery and therefore justifies earlier sampling, while still recording the substantial parts she can now reconstruct independently.

All six have repeated repair history. None should receive the same future programme.

36. A 90-Minute New-Baseline Session

0–8 minutes: fresh independent retrieval with no reminder of the old repair route.
8–16 minutes: inspect what the learner notices, chooses and checks before feedback.
16–24 minutes: changed surface or representation.
24–32 minutes: mixed discrimination against a plausible alternative method.
32–40 minutes: learner explains the repaired principle and the known failure condition.
40–50 minutes: current curriculum task where the capability is actually needed.
50–60 minutes: realistic load only if load belongs to the known fragility.
60–68 minutes: error recovery with tutor support delayed.
68–75 minutes: unrelated work creates a short spacing gap.
75–82 minutes: return to the target without warning.
82–87 minutes: compare evidence with the first and second repair records.
87–90 minutes: write the provisional new baseline and one future falsification test.

The purpose of the session is not to produce a perfect score. It is to expose enough of the system to write a useful current model.

37. The New Baseline Should Survive a Week

A one-week return is useful because it removes the immediate warmth of the repair.

Do not show the same questions. Do not announce the exact test. Put the capability back into ordinary work.

Can the learner retrieve the route?
Can they identify when it is relevant?
Can they resist a competing wrong route?
Can they self-correct?
How much prompting is needed?

If the provisional baseline predicts the learner accurately, confidence in the model increases.

38. The New Baseline Should Survive a Month

A month changes the question.

Now the tutor can ask whether ordinary curriculum, other subjects, school assessments and daily life have displaced the repaired capability from immediate attention.

The learner does not need flawless performance. The tutor is looking for whether the capability remains within the expected working range and whether recovery from a small slip is economical.

A model that remains useful after ordinary life has happened is more valuable than a model written only inside the repair week.

39. Maintenance Should Become Smarter, Not Necessarily Bigger

Repeated repair may justify maintenance.

But maintenance should target the known failure mode.

Fragility	Useful maintenance
Retrieval after low use	Short spaced generation
Method selection	Mixed contrast pairs
Transfer	Occasional changed-surface tasks
Timed execution	Brief realistic timed blocks
Self-correction	Error-finding before tutor feedback
Study initiation	Learner-owned trigger and review routine

The goal is to keep the system observable with the lowest sensible intervention cost.

40. Review Frequency Should Follow Risk and Recovery Cost

Not every known fragility deserves the same monitoring frequency.

A capability that weakened once after a major transition and recovered in ten minutes may need only occasional sampling.

A capability that has reopened twice in ordinary conditions, creates large downstream losses and takes weeks to rebuild deserves closer observation.

Useful factors include:

how often the weakness has reappeared;
how costly failure is;
how early drift can be detected;
how quickly the learner can recover;
how important the capability is to upcoming work;
whether the learner can self-monitor.

41. Primary English Example: Planning Has Reopened Twice

A learner once wrote compositions as disconnected event lists. The first repair built causal planning. Months later, timed assessment pressure produced list-like writing again. A second repair compressed planning into a fast three-anchor route.

The new baseline should not say, “Weak at composition planning.”

It should say that causal planning is understood and available, but speed can strip the plan down too far when the learner begins writing before securing the turning point and consequence.

Future maintenance becomes one three-minute plan every few weeks and one check inside realistic writing—not a return to weeks of basic planning lessons.

42. Primary Mathematics Example: Fractions Have Reopened Twice

The learner understands fraction meaning and equivalent forms. The first recurrence appeared inside percentage. The second appeared inside ratio and rate.

The repeated pattern suggests that the fragility may not be fraction meaning at all. It may be representation choice when several equivalent forms are available.

The new baseline therefore says:

Fraction concepts are stable; representation selection becomes less efficient when fraction, ratio, decimal and percentage compete inside multi-step problems.

That model changes the next route completely.

43. PSLE Science Example: Explanation Structure Has Reopened Twice

The learner knows the route from evidence to concept to mechanism to exact answer. Yet complex investigation questions have twice pulled the student back toward descriptive answers.

The second repair shows that the framework itself is remembered. The breakdown occurs when several pieces of evidence compete for attention.

The new baseline should therefore record strong conceptual route knowledge with a coordination vulnerability under multi-variable evidence load.

Maintenance can now sample complex evidence integration rather than endlessly reteaching the four-step explanation framework.

44. Secondary English Example: Inference Scope Has Reopened Twice

Ciara understands the rule: infer no more than the text can carry. She applies it accurately in short passages. Twice, long timed papers have produced broad unsupported claims.

Her new baseline becomes conditional:

Inference judgement is strong in ordinary reading; evidence anchoring becomes less reliable when passage length and time pressure combine late in a paper.

This description protects what she can do while giving future performance training a precise job.

45. Secondary Mathematics Example: Method Selection Has Reopened Twice

A learner executes algebraic methods well in topical work. Mixed school papers have twice exposed poor method selection.

The new baseline separates execution and selection.

Execution: strong when route is known.
Selection: improving but still vulnerable when similar structures compete.
Self-correction: good after the learner is asked to reclassify the question.
Transfer: acceptable across changed numbers and wording.

That model justifies mixed discrimination practice. It does not justify repeating every algebra chapter.

46. Additional Mathematics Example: Algebraic Control Has Reopened Twice Inside Higher Work

An Additional Mathematics student is accurate on clean algebra diagnostics but has twice lost control during long calculus and trigonometric chains.

The correct baseline is not “weak algebra”.

It may be:

Algebraic knowledge and isolated execution are strong; symbolic compression becomes a reliability risk when several transformations are nested and the learner skips written structure.

The future route therefore protects written structure at transition points rather than sending the learner back to beginner algebra.

47. JC Example: A Mature Learner Can Help Write the Baseline

Older students should increasingly participate in the learner model.

Ask the JC learner:

What part of this capability now feels ordinary?
What condition still changes your reliability?
What is your first self-repair move?
What evidence would convince you that this fragility is no longer important?
What maintenance can you own without the tutor?

The goal is not for the learner to self-diagnose perfectly. It is to make self-observation part of independence.

48. Vocabulary Example: Retrieval and Use Are Different Baselines

A learner may remember definitions while repeatedly failing to retrieve precise words during writing.

Or the learner may retrieve words quickly but misuse register and collocation.

If vocabulary has needed repair twice, the baseline should specify which layer is fragile:

meaning;
distinction from near-synonyms;
free retrieval;
context fit;
register;
spelling;
long-delay access.

“Vocabulary weak” is too large a label to guide a good tutor.

49. Homework Example: Independence Has Reopened Twice

A learner once needed heavy adult control to start and organise homework. Independence improved. Twice, busy periods have brought the old dependence back.

The new baseline should distinguish academic capability from study control.

The learner may be fully capable of the work while initiation becomes fragile when workload exceeds a certain level.

That suggests a learner-owned overload protocol: identify the first task, limit planning time, break the queue into visible actions and escalate only when the workload is genuinely impossible.

The tutor should not reclaim the whole planning system every time pressure rises.

50. Parent Communication: The Child Has a New Map, Not a New Label

Parents deserve a clear explanation when the same weak link has needed repair twice.

We have enough evidence to update how we monitor this area. The capability is real and much of it is now strong, but we have seen that a particular condition can make it less reliable. We are recording that condition, keeping a light check on it and teaching your child how to recognise and repair it independently. We are not restarting the subject, and we are not defining the child by this weakness.

This is more accurate than either reassurance without evidence or alarm without proportion.

51. Student Communication: Make the Model Usable

The student does not need a dossier of educational terminology.

They need a usable rule.

You know this. The part to watch is choosing the method when the topic label disappears. Your first move is to classify the relationship before calculating.

Or:

Your inference judgement is good. Late in long papers, you sometimes stop checking the evidence. Your reset is to anchor one phrase before you write the claim.

A baseline becomes educationally powerful when the learner can act on it.

52. Failure Mode: The Stale Baseline

The tutor keeps using an old description because it was once true.

The learner has changed, the curriculum has changed and the evidence has changed, but the note still says “weak in fractions”, “careless”, “slow reader” or “needs prompting”.

A stale baseline turns history into a ceiling.

If the learner has outgrown the description, retire it.

53. Failure Mode: The Worst-Day Baseline

A painful result becomes the entire model.

One disastrous paper can contain important information. It can expose a load boundary, a missing prerequisite or a performance failure.

But unless the condition is representative, it should not automatically become the new ordinary state.

Use the bad day to ask better questions. Do not let it erase every other receipt.

54. Failure Mode: The Best-Day Baseline

The learner performs exceptionally after focused preparation and adults assume that level is now ordinary.

Then later performance looks like decline even when the learner is operating within a healthy working range.

Peaks are useful because they show possible capability.

Baselines are useful because they show what can be expected more routinely.

Do not punish a learner for failing to live permanently at the peak.

55. Failure Mode: Score-Only Baseline

Marks rise, so the tutor assumes the system is fixed.

But perhaps the learner used heavy support, remembered the exact practice set or avoided the fragile decision because the paper happened not to test it.

Marks matter. Mechanisms matter too.

A useful new baseline combines outcome evidence with evidence about independence, selection, transfer, delay and recovery.

56. Failure Mode: Permanent Watch Mode

Two repairs make the tutor nervous, so the fragility is tested every lesson forever.

This can waste time and keep the learner psychologically attached to an old weakness.

Monitoring should reduce when evidence supports reduction.

A known fragility can move from active repair to maintenance, from maintenance to occasional sampling, and eventually to ordinary background observation.

The baseline is allowed to become more optimistic when reality earns it.

57. Failure Mode: Treating Recurrence as Moral Failure

“You should know this already.”

Sometimes that sentence is factually true and educationally useless.

The tutor’s job is to find out what changed between knowing and not producing: access, selection, transfer, execution, confidence, study routine, representation or load.

Repeated repair should increase precision, not shame.

58. Failure Mode: Lowering Expectations Too Far

A known fragility can make adults overprotective.

The learner receives easier questions, more hints, longer time and fewer opportunities to meet the real condition that once caused the problem.

This can stabilise performance by removing the world the learner must eventually face.

The new baseline should help dose challenge accurately. It should not become an excuse to permanently lower the ceiling.

59. The New-Baseline Ledger

Field	Record
Capability	What can the learner now do correctly and independently?
Working range	What is ordinary under realistic conditions?
Floor	What survives weaker days or mild disruption?
Ceiling	What appears when conditions align well?
Known fragility	Which repeatable condition has weakened the capability?
First repair mechanism	What originally needed change?
Second repair mechanism	What needed change the second time?
Recovery cost	How much help and time were needed?
Support demand	What level of prompting remains?
Selection	Can the learner choose the route without labels?
Transfer	Across which changed surfaces does the capability survive?
Delay	Does it remain accessible after time?
Performance load	What happens under timing, length and mixed work?
Self-correction	Can the learner detect and repair slips?
Study control	Can the learner maintain the capability between lessons?
Education context	Which current or upcoming curriculum demands matter?
Maintenance dose	What is the smallest useful future check?
Review trigger	What evidence would justify earlier re-checking?
Retirement condition	What evidence would allow the fragility watch to shrink or disappear?

60. The New-Baseline Decision Tree

Step 1: Confirm that the second repair has produced enough independent evidence to update the learner model.
Step 2: Separate current capability from known fragility.
Step 3: Describe the ordinary working range rather than using a peak or worst day.
Step 4: Record the condition under which the weakness has reopened.
Step 5: Compare first-repair and second-repair mechanisms.
Step 6: Record recovery cost and support demand.
Step 7: Test selection, transfer, delay and self-correction.
Step 8: Add realistic load when load is relevant.
Step 9: Decide whether learning, studying, training, education route or improvement loop needs to change.
Step 10: Choose the smallest maintenance dose justified by the fragility.
Step 11: Give the learner one usable self-monitoring rule.
Step 12: Explain the update to parents without converting fragility into identity.
Step 13: Schedule a delayed falsification test.
Step 14: Reduce monitoring when later evidence supports reduction.
Step 15: Rewrite the baseline again when reality changes.

61. What Research Supports

The logic of a new baseline draws on several well-established ideas without claiming that research provides one universal “post-second-repair baseline protocol”.

Research distinguishing learning from immediate performance supports caution about treating one supported or recent success as durable capability. Work on retrieval practice and spacing supports the use of delayed generation rather than repeated exposure alone. Research on transfer supports checking whether learning survives changes in surface or context. Work on metacognition and calibration supports comparing confidence with actual performance rather than relying only on a learner’s feeling of knowing. Educational measurement more broadly supports using multiple observations and comparable conditions when making stronger claims about change.

62. What Research Does Not Justify

Research does not justify declaring a child permanently fragile because the same weakness has required two repairs.

It does not justify a fixed monitoring schedule for every learner.

It does not justify treating every drop in marks as evidence that the repaired capability has disappeared.

It does not justify endless retrieval practice when the real problem is selection, transfer or execution.

It does not justify lowering task difficulty forever in the name of confidence.

It does not justify confusing a tutor’s neat model with the learner’s full reality.

63. The Model Must Be Easy to Revise

A learner model that is too elaborate can become hard to change because adults grow attached to the paperwork.

Keep the important parts visible:

what is strong;
what is conditionally fragile;
what evidence supports that claim;
what the learner can do to self-correct;
what future condition deserves a check;
what evidence would allow the model to become more optimistic.

The model should serve teaching. Teaching should not become a servant of the model.

64. The New-Baseline Operating Cycle

Complete the second repair → collect fresh independent evidence → delay and return → change the surface → mix competing routes → add realistic load where relevant → inspect support demand → inspect self-correction → compare first and second repair histories → separate capability from fragility → define floor, working range and ceiling → update the mechanism hypothesis → update studying only if study control needs change → update training only at the vulnerable boundary → update the education route around real upcoming demands → choose proportionate maintenance → give the learner a self-monitoring rule → explain the model without labelling the child → schedule a later falsification test → reduce monitoring when evidence allows → rewrite the baseline whenever reality changes.

Evidence and Further Reading

Final Compression

The learner had a weak link.

It was repaired.

It returned.

It was repaired again.

Now the tutor has more than a repaired skill.

The tutor has history.

Use that history carefully.

Do not erase the learner’s strength because one part of the system has shown fragility. Do not erase the fragility because the learner has become strong again. Hold both truths at once.

Record what the learner can do independently. Record the conditions that have previously weakened it. Record how much support recovery required. Record whether the learner can now detect and repair the problem earlier. Separate ordinary working range from peak performance and from the occasional bad day.

Then let the model change the right layer.

If the issue is learning, strengthen capability. If it is studying, repair the learner’s deliberate routine. If it is training, practise the boundary. If it is education, adjust the route around current demands. If it is improvement, choose the next testable change.

Keep maintenance proportionate.

Keep the learner involved.

Keep the hypothesis falsifiable.

And keep rewriting the map when reality changes.

A strong tutor does not preserve an old learner model because it is familiar, and does not replace it with a harsher label because a weakness returned. The tutor builds a more precise forecast: capability here, fragility there, recovery route known, independence increasing, future evidence still allowed to change the story.

That is the discipline of The New Baseline.

That is Tutor Handbook Volume 0031.

Next in the series: The Tutor Handbook Vol No.0032 | The Watchlist — How a Tutor Monitors a Known Fragility Without Making It the Centre of Every Lesson.

Small Group Tutorials