The Tutor Handbook

Canonical root · Tutor Handbook operating manual · live series through Vol. 0210

The Tutor Handbook is not a collection of tips arranged in numerical order. It is a decision system for tutoring.

A tutor sits beside a learner and sees a problem: an unfinished homework question, a wrong answer, a slow response, a sudden mark drop, a strong result that may not hold, a student who can follow an example but cannot begin alone, a group in which one learner understands while two appear to understand, a parent who wants faster progress, a school method that conflicts with the tuition method, or a teaching routine that looked promising in training but does not survive ordinary lessons.

The visible event is not yet the teaching job.

The tutor's real work is to decide what the event means, what evidence is missing, what action is justified, how much support is appropriate, what result would count as improvement, when the route should change, and when tuition should stop doing the work because the learner can now carry it.

That is why this Handbook exists.

It gives a name and a boundary to recurring tutoring decisions. Each volume owns one decision job. The root page you are reading owns the operating architecture that connects those jobs without allowing them to collapse into one giant method.

This matters because good tutoring is not simply "more teaching". It is not a smaller classroom by itself. It is not the possession of more worksheets. It is not a warm relationship by itself, a high examination score by itself, a charismatic explanation by itself, a sophisticated AI tool by itself, or a long list of evidence-informed techniques by itself.

Good tutoring is a controlled sequence of professional decisions made close to the learner.

A useful shorthand for the entire Handbook is:

observe → define the job → collect enough evidence → separate competing explanations → choose the smallest useful intervention → teach → transfer responsibility → verify on fresh evidence → review → continue, change, release or escalate

Every volume in the Tutor Handbook should be readable as one part of that loop.

Why the Handbook needs an operating manual

A long series creates a new problem as it becomes successful.

At ten articles, a reader can browse.

At fifty articles, the reader needs categories.

At one hundred articles, categories alone become insufficient because several articles will appear relevant to the same event.

At two hundred articles, the series needs an explicit operating model. Otherwise the reader can find information yet still not know which page owns the next decision.

Consider a learner who answers a Mathematics question correctly after looking at a worked example.

Several Handbook jobs may be nearby.

The Correct-Answer Audit can ask whether the right answer actually reflects sound reasoning.

The Support Provenance Check can ask what help contributed to the answer.

The Evidence Freshness Window can ask whether the success occurred too close to teaching or prompting to count as independent learning.

The Worked-Example-to-Problem Alternation Gate can ask when the learner should move from studying the worked solution to carrying more of the route.

The Fade can ask how support should be removed.

The Independence Test can ask whether unsupported performance is stable enough to count as self-regulation.

The Changed Question can ask whether the learning survives a surface change.

The Mixed Set can ask whether the learner can select the method when no topic label announces it.

These are not duplicates if each article keeps its job.

The same event can legitimately pass through several gates because the gates answer different questions. The root manual therefore controls sequence and ownership, not merely classification.

A Handbook article earns its place when removing it would leave a distinct decision unanswered.

The central object is the learning route

The Handbook uses many words—lesson, method, intervention, support, routine, practice, check, gate, boundary, differential, audit, review—but the system becomes easier to understand when one object sits at the centre: the learning route.

A learning route is the current working plan that links:

• the capability being built,
• the learner's present evidence,
• the tutor's best current explanation of the bottleneck,
• the teaching or practice being used,
• the support conditions,
• the evidence that will be collected next,
• the threshold for continuing or changing course,
• and the route towards independence.

A route can be small. It might concern one misconception in fractions.

A route can be wider. It might concern six weeks of examination preparation.

A route can be individual. It might exist for one learner inside a three-student tutorial.

A route can also exist at programme level. A tuition centre might adopt a new feedback routine, tutor-observation system or AI-material verification procedure and then need to know whether it is being implemented consistently.

The Handbook should not treat a route as a script. A script specifies what happens regardless of new evidence. A route specifies the current best plan and the conditions under which that plan should change.

That distinction is one of the deepest principles in the series.

Preparation is necessary.

Pre-deciding the learner is not.

Consistency is useful.

Rigidity is not.

Adaptation is necessary.

Uncontrolled improvisation is not.

The Handbook lives in those boundaries.

Tutoring is a sequence of state changes, not a pile of activities

A lesson can contain explanation, practice, feedback, questioning, review and homework, yet none of those activities is automatically useful.

The useful question is: what state is the learner in before the activity, and what state should be different after it?

If the learner does not understand a concept, explanation may be appropriate.

If the learner understands but cannot execute accurately, practice may be appropriate.

If execution is accurate but slow, fluency work may be appropriate.

If the learner performs only with prompts, support reduction may be appropriate.

If the learner succeeds on rehearsed questions but fails on fresh ones, transfer work may be appropriate.

If the learner performs well alone but poorly under time pressure, performance coaching may be appropriate.

If the learner's school marks drop while tuition work looks strong, the problem may be evidence comparability, task conditions, exam technique, curriculum mismatch or a hidden weakness outside the sampled work.

The tutor is therefore not choosing activities in a vacuum. The tutor is attempting to move the learner from one defensible state description to another.

This is why so many later Handbook articles are called gates, checks, boundaries, differentials and audits.

A gate asks whether the conditions for a move are satisfied.

A check asks whether an assumption or result survives inspection.

A boundary prevents one legitimate activity from expanding into a different and potentially harmful job.

A differential separates plausible causes that require different responses.

An audit reopens something that looks settled and asks whether the evidence actually supports the conclusion.

A window defines how much or how recent evidence should be considered.

A protocol controls what happens when people or evidence disagree.

These are different control devices inside one tutoring operating system.

The Handbook's basic unit: one article, one decision job

Every Tutor Handbook article should own one bounded professional decision.

That job should be expressible in one sentence beginning with a verb:

• decide whether,
• distinguish between,
• determine when,
• choose how much,
• verify that,
• interpret whether,
• preserve enough,
• remove support without,
• coordinate without,
• report without,
• adapt without,
• implement without,
• or escalate when.

The article is not commissioned because a keyword exists. It is commissioned because a tutor, coach, programme lead, learner or family repeatedly reaches a decision point that deserves an owner.

The article should answer one primary question completely enough that another Handbook page does not need to reproduce the same mechanism.

This does not make the articles narrow in value. A bounded decision can require deep treatment.

The Tutor Selection Evidence Gate, for example, can require discussion of credentials, demonstrations, observation, evidence samples, subject knowledge, instructional judgement, safeguarding boundaries, structured selection and the limitations of a single polished trial lesson. The article can be long while its owner job remains precise.

The Implementation Sustainability Gate can examine staffing, routine burden, leadership dependency, material maintenance, observation, training, feedback and long-term use. It remains one job: deciding whether a practice can remain alive under ordinary conditions rather than only during launch.

Length follows functional completeness, not topic inflation.

That principle controls this root page too. This manual is long because it must define the architecture that allows more than two hundred decision owners to remain coherent.

The Tutor Handbook job contract

From this point forward, every Tutor Handbook article should satisfy the same canonical contract.

1. Trigger

What event makes this article relevant?

A trigger must be observable enough that a tutor can recognise it. "The learner is weak" is too vague. "The learner can solve near-copy examples but cannot begin a minimally changed problem" is a useful trigger.

2. Decision

What specific professional decision does the article own?

If two articles claim the same decision, the collision must be repaired before another article is published.

3. Evidence input

What evidence should the tutor inspect before acting?

The answer can include live work, a fresh item, learner explanation, school results, timing data, support history, parent observations, attendance, lesson recordings, curriculum materials or programme implementation evidence.

The article must distinguish useful evidence from signals that are vivid but not decisive.

4. Competing explanations

What else could produce the same visible symptom?

A low score may come from missing knowledge, reading load, timing, careless execution, unfamiliar format, anxiety, fatigue, weak method selection, or a task that sampled the wrong construct.

A late-session decline may be fatigue, but it may also be rising difficulty, accumulated error, reduced support, a change of topic or simple random variation.

A Handbook article should make the tutor earn the diagnosis.

5. Permitted action

What may the tutor reasonably do if the evidence supports the interpretation?

The action should be proportionate. The Handbook prefers the smallest useful intervention because large interventions change more variables, cost more time and make the next result harder to interpret.

6. Support condition

How much help is allowed while the target capability is being built or measured?

Many tutoring errors are support errors. A learner appears competent because the tutor, parent, peer, worked example, AI system or worksheet structure silently carries part of the target skill.

The article should state when support is instructional and when support would contaminate the evidence.

7. Receipt

What observable result shows that the intervention did its job?

Agreement is not a receipt.

Enjoyment is not a receipt.

Completion is not necessarily a receipt.

A receipt is evidence tied to the target decision: a changed next attempt, a fresh independent performance, a stable teaching move observed in live practice, a route that remains coherent after handover, or an implementation practice that continues under ordinary workload.

8. Stop, hold or change rule

When should the current route continue? When should it be held steady? When should it change? When should it end?

A route with no exit rule can become permanent merely because nobody knows when to stop.

9. Failure modes

What predictable mistakes make the method look better or worse than it is?

Every Handbook article should include failure modes because tutoring decisions fail in patterned ways.

10. Ownership boundary

Which neighbouring Handbook jobs are related but distinct?

This protects the series from cannibalisation and helps the reader move to the next correct owner.

11. Evidence boundary

What can research, guidance or local evidence legitimately support—and what remains professional judgement?

The Handbook should not convert research averages into guarantees for the learner in front of the tutor.

12. Revision rule

What new evidence would cause this article's recommendation to be reconsidered?

A professional handbook should be stable enough to use and open enough to improve.

That twelve-part contract is the job grammar of the series.

A control loop for live tutoring

The following loop can be used when a tutor does not know which Handbook article to open.

Observe

Start with what happened, not with the label.

Write the event in neutral language.

"Three fraction questions were wrong after the denominator changed."

"She answered correctly after I pointed to the relevant sentence."

"He completed the paper but left the final two questions blank."

"The group produced one correct explanation after discussion."

"The new tutor can describe the feedback routine but did not use it in the observed lesson."

This protects the first step from premature diagnosis.

Define the target

What capability is actually being taught or measured?

If the target is inference, reading speed may be relevant but is not the same skill.

If the target is algebraic method selection, arithmetic accuracy can obscure the result.

If the target is independent writing, a model paragraph that supplies the structure may be legitimate during teaching but not during verification.

A tutor cannot interpret evidence before defining what the evidence is supposed to represent.

Identify the first consequential uncertainty

Do not ask every possible question at once.

Find the uncertainty that, if resolved, would most change the next action.

Does the learner know the concept but fail to select the method?

Was the homework independent?

Did the low mark come from one component?

Did the learner misunderstand the task or lack the prerequisite?

Did the tutor actually implement the routine that is being evaluated?

This is the moment when a specific Handbook owner should become visible.

Collect the smallest discriminating evidence

More data is not automatically better data.

A single well-designed diagnostic probe can sometimes separate two competing explanations better than twenty routine questions.

One fresh item can reveal whether a worked example was copied or understood.

One short delayed return can reveal whether a repair held.

One observed live lesson can reveal whether training transferred—provided the observation sample is not mistaken for normal practice without further evidence.

The Handbook repeatedly favours evidence that changes a decision.

Act proportionately

Choose the smallest action likely to test or improve the route.

This may be one explanation, one contrast case, one partially worked example, one changed question, one timed set, one regrouping, one coaching focus, or one temporary implementation trial.

Large-scale change should require a larger evidence burden because it creates larger costs and more confounding.

Transfer responsibility

Tutoring that never transfers responsibility can produce impressive sessions and weak learners.

The tutor should ask: what part of this performance can the learner now own?

Support can move from full demonstration to guided completion, from prompts to delayed prompts, from visible examples to contingent access, from tutor-selected practice to learner-selected practice with verification, or from intensive sessions to check-ins.

The correct amount of support is not "as little as possible". It is enough to make productive learning possible while preserving a route towards independence.

Verify under a changed condition

Immediate success after teaching is useful but limited.

Verification becomes stronger when one relevant condition changes:

• time passes,
• the question surface changes,
• method labels disappear,
• the example is hidden,
• the tutor reduces prompting,
• the learner explains the decision,
• the task is mixed with alternatives,
• the work occurs outside the lesson,
• or a second tutor receives the route.

Freshness is not an aesthetic preference. It is part of determining what the learner can actually carry.

Review the route

The review asks a decision question, not merely "How are things going?"

Continue?

Advance?

Change method?

Repair prerequisite?

Increase or decrease support?

Regroup?

Escalate?

Release?

Collect more evidence?

The answer should be tied to the evidence that was specified before the review whenever possible.

This prevents the rule from moving after the result arrives.

The root manual does not replace the volumes

The purpose of this page is orientation, architecture and control.

It should not become a duplicate of all child articles.

When this manual explains a principle such as support fading, it explains why that principle exists in the system and points to the relevant owner for operational depth.

When it discusses evidence freshness, it defines the role of freshness in the wider loop rather than reprinting the entire Evidence Freshness Window.

When it discusses programme implementation, it explains why tutor training, feasibility, acceptability, reach, consistency and sustainability require separate owners rather than compressing them into one implementation chapter.

The root is the map and the operating doctrine.

The volumes are the instruments.

The first weak link is not always the first visible error

Tutoring becomes inefficient when the tutor treats the latest mistake as the whole problem.

A learner may make an arithmetic error at the end of a long algebra question. Correcting the arithmetic is necessary, but it may not be the weak link that controls performance. The learner may have selected the wrong algebraic structure earlier and arrived at a calculation that should never have existed.

A student may produce a weak comprehension answer because the final sentence is vague. The deeper issue may be that the evidence chosen from the passage does not support the claim.

A Science learner may omit a key term. The omission may be vocabulary, or it may reveal that the underlying causal mechanism is not understood.

A writer may produce repetitive sentences. The problem may be language control, but it may instead be an undeveloped idea that gives the sentence nothing useful to express.

The Handbook's diagnostic articles exist to stop the tutor from confusing the location of an error with the cause of the performance.

A weak link is consequential when repairing it changes what becomes possible next.

If a learner cannot decode the question accurately, method instruction may not transfer.

If prerequisite number sense is unstable, a new algebra technique may be memorised rather than understood.

If the learner knows the method but cannot recognise when to use it, more blocked practice may improve speed without improving method selection.

The tutor should therefore ask:

1. What failed?
2. What had to be true before that step could succeed?
3. Which earlier condition is both weak and consequential?
4. What evidence would separate that explanation from plausible alternatives?

This is the job behind the Diagnostic Tutor, Fresh Look, Differential, Diagnostic Probe, Disconfirmation Check, Tutor-Side Check, Task Purity Check, Prerequisite Repair Boundary and related owners.

They do not all diagnose the same thing.

The Diagnostic Tutor establishes the broad function.

The Fresh Look protects against allowing old labels to determine a new case.

The Differential separates competing causes.

The Diagnostic Probe designs an efficient test between them.

The Disconfirmation Check asks the tutor to look actively for evidence that would prove the current explanation wrong.

The Tutor-Side Check asks whether the tutor's own explanation, sequence or prompt created the difficulty.

The Task Purity Check asks whether the task contains hidden demands that contaminate interpretation.

The Prerequisite Repair Boundary asks how far backwards the tutor should go once an earlier dependency is identified.

The root manual sequences those jobs.

Use hypotheses, not identities

A useful tutor diagnosis is provisional and actionable.

"The learner may be failing to identify the denominator that defines the comparison" is useful.

"The learner is weak at percentages" may be too broad.

"The learner is careless" is often worse because it turns an observed error pattern into a personal trait.

"She is not a Maths person" is not a professional learning diagnosis at all.

A hypothesis should point to evidence and a possible next test.

This is why the Handbook prefers phrases such as:

• current evidence suggests,
• one plausible explanation is,
• this result is consistent with,
• the next task should separate,
• the route should remain provisional until,
• and the present evidence does not yet distinguish.

Provisional language is not indecision. It is disciplined uncertainty.

The Provisional Language Gate applies this especially to communication with parents. Families deserve clarity, but clarity is not the same as false certainty.

A tutor can say:

"Across three recent tasks, he selects the correct formula when the topic is named, but performance drops when several methods are mixed. I am treating method selection as the current bottleneck. We will test that with fresh mixed questions before changing the teaching route."

That is more useful than either extreme:

"He doesn't understand the topic."

or

"We can't know anything yet."

Professional judgement lives between those extremes.

Evidence has conditions

A score does not arrive alone. It arrives with a condition.

Who selected the questions?

Was the work timed?

Was the method announced?

Were examples visible?

Did the tutor prompt?

Did a peer speak first?

Was AI used?

Was the work completed immediately after teaching or several days later?

Was the same question seen before?

Did the learner have the same access accommodations?

Was the task easier or harder than the previous one?

Was the standard of marking stable?

Was the learner observed under ordinary conditions?

Did everyone who started the programme appear in the final data?

The Handbook contains many later articles because these conditions determine what an apparently simple result can mean.

The Support Provenance Check asks what assistance contributed to the output.

The Evidence Freshness Window asks how close success occurred to teaching and rehearsal.

The Accommodation Baseline asks whether legitimate access supports were stable enough for comparisons.

The Measurement Range Check asks whether the task was capable of showing meaningful improvement.

The Criterion Drift Check asks whether the standard quietly changed.

The Adaptive-Difficulty Check and Adaptive-Difficulty Comparability Gate ask what happens when the platform changes question difficulty.

The Attrition Evidence Gap asks what it means when learners who stop attending disappear from final outcomes.

The Observation Reactivity Check, Observation-Sample Gate and Tutor-Observation Sample Gate ask whether observed lessons represent ordinary practice.

The goal is not to make measurement impossible.

The goal is to stop numbers from claiming more than their conditions allow.

Do not average unlike evidence into a fake certainty

Tutors often possess several sources of information:

a school examination,

a tuition worksheet,

a learner's self-report,

a parent's observation,

a teacher comment,

a timed practice set,

a tutor's live observation,

and perhaps an AI-generated practice history.

The temptation is to combine them into one intuitive average: "Overall, she is about 70% secure."

That number may have no coherent meaning.

Different evidence can answer different questions.

A school exam may reveal performance under standardised time and format.

A tuition worksheet may reveal performance with recent teaching and known task types.

A learner report may reveal confidence, perceived difficulty or strategy use.

A parent may observe homework behaviour unavailable to the tutor.

A live tutoring session may reveal reasoning before the final answer is written.

The Evidence Triangulation Check does not ask the tutor to average unlike sources. It asks whether several differently conditioned sources converge on the same interpretation.

Triangulation is strongest when the evidence sources have different weaknesses.

Three near-identical worksheets are not three independent confirmations merely because they were completed on different days. That is the logic behind the Correlated-Evidence Trap.

The tutor wants evidence diversity where diversity helps answer the decision.

A correct answer can still contain uncertainty

Correctness is an outcome, not always an explanation.

A learner can arrive at the right answer by:

guessing,

copying a pattern,

following a peer,

using an invalid method that happens to work on one case,

receiving a prompt,

selecting from limited options,

or applying a memorised procedure without recognising its boundary.

This does not mean every correct answer should trigger interrogation.

That would make tutoring exhausting and would turn success into suspicion.

The Correct-Answer Audit exists for moments when the cost of false confidence is high or the success conflicts with other evidence.

A tutor might audit when:

• the answer arrives unusually quickly,
• the method is hidden,
• a known misconception could produce the same answer,
• the learner previously failed structurally similar tasks,
• support conditions changed,
• or the task is too narrow to establish the broader capability.

The audit can be light.

"How did you decide?"

"Would the same method work here?"

"Show me the first step on this changed version."

A small probe can protect a large future decision.

Wrong answers also require interpretation

The same wrong answer can arise from different causes.

A blank response can mean:

no knowledge,

uncertainty,

time shortage,

avoidance,

failure to notice the question,

difficulty reading the wording,

a deliberate skip-and-return strategy,

or interrupted work.

The Omission Audit and Missing-Response Differential exist because "blank = does not know" is too crude.

A slow correct response can mean fragile recall, careful verification, a deliberate strategy, language processing demand, unfamiliar format, or a learner who understands deeply but has not yet automated execution.

The Response-Time Interpretation Gate prevents "fast = smart" and "slow = weak" from becoming invisible assumptions.

An off-task learner may be disengaged, or the task may be too difficult, too easy, inaccessible, badly sequenced, socially uncomfortable or already understood.

The Behaviour–Learning Differential and Participation Differential protect instruction from moralising a problem that may be instructional.

Comparison requires a defensible baseline

Progress means change relative to something.

But the baseline may be missing, unstable or incomparable.

The Missing Baseline asks what can honestly be said when tutoring began before a clean starting measure was collected.

The New Baseline asks when the learner model itself should be rebuilt after repeated repair.

The Peer Baseline Trap stops the fastest learner in a group from becoming the implicit standard for everyone.

The Programme-Average Boundary stops an average tutoring outcome from pretending to describe the individual learner.

The Learner-Progress Data Ownership Gate later asks how evidence can be shared with learners so they can use it without turning scores into identity.

A baseline should serve a decision.

It should not become a permanent label.

If the learner changes substantially, the most useful baseline for a new phase may be the learner's recent stable performance, not the mark from the first week of tuition.

Look for trend, but do not worship the line

The Handbook's early progress sequence—Trend, Plateau, Breakthrough, Maintenance, Reopening, Second Repair, New Baseline, Watchlist, Retirement, Archive—treats learning as history rather than a string of isolated marks.

A trend can be useful when tasks are sufficiently comparable and the direction persists.

A plateau can signal stalled learning, but it can also signal that the measurement has reached a ceiling, that the learner is working on harder material, or that gains shifted to a capability the score does not capture.

A breakthrough can represent real phase change, but it can also be one favourable paper.

Maintenance exists because improvement that disappears after intervention ends was not yet stable enough for release.

Reopening exists because a repaired weakness can return under new conditions.

The Watchlist prevents a known fragility from consuming every lesson.

Retirement prevents permanent surveillance of an old problem that no longer deserves special attention.

The Archive preserves useful history while refusing to let history become destiny.

The sequence embodies a principle: learning evidence has memory, but professional judgement must remain updateable.

Separate improvement from causal claims

Suppose a learner's marks rise from 58 to 72 after six weeks of tuition.

The improvement is real as an observed difference if the scores are comparable.

It does not automatically prove that tuition caused fourteen marks of gain.

School teaching continued.

The learner practised independently.

The examination content may have differed.

Regression to the mean may matter if the first score was unusually low.

The second paper may be easier.

The learner may have slept better.

The tuition may indeed have contributed substantially, but the evidence may not isolate that contribution.

The Causal-Attribution Gap protects reporting.

The Regression-to-the-Mean Trap protects against celebrating a rebound after an extreme result as though the intervention necessarily produced it.

The Aggregate-Score Compensation Trap asks whether a total score hides a serious component weakness.

The Subscore-Resolution Trap asks whether tiny diagnostic categories pretend to be more precise than the number of items allows.

These articles matter because professional credibility depends on saying only what the evidence supports.

A tutor can still be useful without claiming experimental certainty.

"We observed a sustained improvement across three fresh papers after the route changed, including the component we targeted. That is encouraging evidence that the current approach is useful. We cannot isolate tuition as the only cause."

That statement is strong because it is proportionate.

Error costs change the evidence burden

Not every decision requires the same amount of evidence.

If the tutor is deciding whether to try one different practice question, the cost of error is small.

If the tutor is deciding to move a learner down an entire curriculum level, change tutors, add several weekly sessions, stop tuition, or tell a family that the learner has mastered a capability, the cost of being wrong is larger.

The Error-Cost Asymmetry Gate asks whether acting too early and waiting too long carry the same cost.

Often they do not.

A tutor may require stronger evidence before making an irreversible or identity-shaping decision.

At the same time, the tutor should not use "we need more evidence" as a reason to delay a low-cost helpful action indefinitely.

The evidence burden should be proportional to the consequence.

This is a general Handbook rule.

Precommit what would change your mind

One of the simplest protections against hindsight is to define a route-change threshold before the next result arrives.

The Decision-Threshold Gate and Precommitted Route-Change Threshold emerged around this problem.

Before the fresh check, state:

• what result would support continuing,
• what result would trigger more evidence,
• what result would justify changing the route,
• and what result would reverse a previous change.

The threshold does not have to be a single score.

It may be a pattern.

"Two fresh independent questions correct using the method for the right reason."

"Three sessions in which the routine is observed under ordinary conditions."

"A second paper showing the repaired component remains above the prior range."

Precommitment makes professional judgement inspectable.

It also makes disagreement easier to resolve because the conversation becomes about the rule and the evidence, not about who feels more confident after seeing the outcome.

Tutoring quality is partly the quality of restraint

A tutor has many ways to help.

That is precisely why restraint matters.

When the learner struggles, the tutor can explain again, simplify the question, give a hint, show a worked example, provide a formula, ask a leading question, remove a distractor, give extra time, let a peer answer first, assign more practice, add a session, change materials, switch methods or take over part of the task.

Every form of help changes the evidence.

The professional question is not "Can I help?"

It is:

What is the smallest support that makes productive learning possible without performing the target skill for the learner?

The Dose owns the amount of help.

The Branch Point owns the smallest intervention after causes have been separated.

The Access-Support Boundary distinguishes legitimate access help from assistance that carries the target capability.

The Question Return Gate asks when to answer, clarify or hand part of the question back.

The Follow-Up Question asks how to respond to a partly correct answer without feeding the route.

The Revoicing Boundary asks how to restate a learner's thinking without quietly upgrading it.

The Criteria Disclosure Gate asks how to make success visible without creating an answer key.

These articles share a philosophy: support should create access to learning, not create artificial evidence of learning.

The intervention should have one primary job

A common mistake is to change many things at once.

The tutor changes the worksheet, explanation style, homework volume, session length, grouping, topic sequence and marking routine in the same week.

The learner improves.

What worked?

The learner declines.

What failed?

The Change Queue, Change Bandwidth, Concurrency Problem and Configuration Snapshot exist because simultaneous change destroys interpretability.

Sometimes several changes are unavoidable. School, family, tuition and the learner can all alter the system at once.

The tutor cannot freeze the world.

But the tutor can keep a record of what changed, limit discretionary changes, and identify the intervention expected to do the main job.

That turns a messy environment into a more readable one.

Design an exit before support becomes normal

The Re-Entry Plan contains a rule that belongs across the Handbook: temporary support should know how it will end.

If a learner receives extra prompting, when will prompting reduce?

If a learner gets two additional sessions, what evidence will return the cadence to normal?

If a tutor temporarily simplifies question language, when will full academic language return?

If a programme introduces intensive coaching for a tutor, what evidence will allow support to taper?

Without an exit condition, an intervention can survive because it is comfortable, not because it is still needed.

The learner then becomes adapted to the support.

The tutor becomes adapted to providing it.

The programme budget becomes adapted to funding it.

The Handbook asks the harder question: has the capability changed enough that the system can now do less?

That is not withdrawal for its own sake.

It is the purpose of successful support.

Explanation is a bridge, not the destination

The first volumes of the Handbook begin with recognisable tutoring roles because those roles remain useful even after the series becomes technically sophisticated.

The Explainer makes difficult ideas clear.

That sounds simple until the tutor asks what "clear" means.

A clear explanation should make the relevant structure easier to perceive, connect new material to what the learner already knows, reduce unnecessary complexity, and leave enough cognitive work for the learner to build the capability rather than merely admire the tutor's competence.

A tutor can explain beautifully and still fail to teach.

The learner may understand while listening yet be unable to reconstruct the idea later.

The learner may imitate the language without understanding the mechanism.

The learner may remember the example but not recognise the idea in a changed representation.

The tutor may answer every question so quickly that uncertainty never becomes visible.

This is why explanation should normally be followed by a receipt.

Ask the learner to:

• restate the relationship in their own words,
• predict a consequence,
• complete one consequential step,
• choose between examples and non-examples,
• solve a paired problem,
• explain why an alternative route fails,
• or use the idea after the representation changes.

The explanation's job is not finished when the tutor stops speaking.

It is finished when the learner can do something newly possible because the structure became available.

Explicit teaching and professional adaptation are not opposites

Evidence-informed guidance on explicit instruction emphasises clear explanation, modelling, sequenced content, guided practice, checks for understanding and movement towards independent practice. The Australian Education Research Organisation's current explicit-instruction practice guide, updated 8 September 2026, describes teaching as structured and sequenced and explicitly discusses worked examples and the gradual removal of scaffolding as proficiency grows.

The Handbook uses those principles without turning them into a universal script.

A tutor works with a much smaller number of learners than a classroom teacher. That creates opportunity for tighter diagnosis and faster adaptation, but it also creates risk: because the tutor can respond instantly to every difficulty, the learner can become over-supported.

The tutor should therefore hold two truths together.

New or complex learning often benefits from clear, structured teaching.

The amount and form of structure should change when learner evidence changes.

The Preparation–Adaptation Boundary, Preparation–Contingency Boundary, Session Structure Gate, Material Override Gate, Instructional Material Coherence Gate and Active-Ingredient Adaptation Gate exist in that space.

Preparation gives the lesson a coherent route.

Contingency prevents the route from overriding fresh evidence.

Structured materials reduce arbitrary variation.

The override gate allows departure when the learner evidence justifies it.

The active-ingredient gate asks whether an adaptation preserved the mechanism that made the practice useful.

Good adaptation is not improvisation without memory.

It is controlled variation around a clear instructional purpose.

Worked examples should reveal decisions

The Handbook article for this worked-example decision is Vol. 0202, The Worked-Example-to-Problem Alternation Gate.

Its presence near the frontier is significant because it captures a central tutoring problem.

A learner can understand a worked example and still be unable to solve the next problem.

That is not paradoxical.

Studying a complete solution and generating a solution are different tasks.

The worked example reduces search. It can make the structure visible. It can model a procedure, a representation, a causal chain or a writing decision.

But if the example carries every consequential choice, the learner may never practise making those choices.

The bridge can include:

• predicting the next step before it is shown,
• explaining why one move is legitimate,
• completing a partially worked example,
• solving a minimally changed paired problem,
• comparing two correct methods,
• identifying the first error in an erroneous example,
• hiding the example after an initial attempt,
• or returning to a fresh problem after a delay.

The wider Handbook surrounds this mechanism with distinct owners.

The Example-Variation Gate asks what should vary and what should remain invariant across examples.

The Self-Explanation Readiness Gate asks when asking "why?" will deepen learning and when the learner first needs more knowledge.

The Erroneous-Example Gate controls deliberately incorrect solutions.

The Contrast-Case Gate controls which examples should sit side by side.

The Worked-Example-to-Problem Alternation Gate controls when solution information should remain visible and when the learner should carry more of the route.

The Fade owns the broader reduction of support.

The Independence Test asks whether unsupported performance is stable.

This is what proper article ownership looks like: related mechanisms, separate decisions.

Practice must match the capability being built

The Drill Builder exists because repetition can build fluency, but only when the repeated activity exercises the right process.

Twenty identical questions can make a procedure faster.

They can also make the learner dependent on a topic label.

A block of practice can tell the learner, "Every question on this page uses the same method." That removes method selection from the task.

The Mixed Set reintroduces selection.

The Interleaving Readiness Gate asks when methods are stable enough to be mixed without producing unproductive confusion.

The Challenge Readiness Gate asks whether harder work will extend the learner or merely hide unstable foundations.

The Fresh-Item Reserve protects some questions from teaching and rehearsal so later evidence still means something.

The Task-Purpose Gate asks whether a question is being used for teaching, practice, diagnosis, progress monitoring or verification. The amount of help allowed depends on that purpose.

This produces a simple but powerful rule:

Do not judge a practice item by whether the learner completed it. Judge it by whether it exercised the capability the route is trying to build.

If the target is recall, retrieval matters.

If the target is method selection, labels must eventually disappear.

If the target is examination timing, the clock must eventually enter.

If the target is transfer, surface conditions must eventually change.

If the target is independent writing, models must eventually stop carrying structure and phrasing.

If the target is collaborative reasoning, each learner still needs an individual accountability condition somewhere in the sequence.

Fluency, understanding and transfer are different receipts

A learner may become fast without understanding.

A learner may understand a concept but execute slowly.

A learner may execute accurately in familiar tasks but fail to recognise the same structure in a new context.

The Handbook therefore separates several performance states.

Understanding concerns the relationships, reasons or meanings that organise the knowledge.

Fluency concerns reliable and efficient execution where speed is genuinely part of useful performance.

Transfer concerns using the capability when the surface, context or task arrangement changes.

Selection concerns choosing the appropriate method when alternatives exist.

Performance under conditions concerns producing the capability under time pressure, examination format or other real constraints.

These states can overlap. They should not be collapsed.

The Changed Question, Mixed Set, Timed Set, Full Paper, Second Full Paper and later measurement owners create a progression from learning inside a supported local task to performance inside a more complete system.

The progression is not always linear.

A learner may need to return from full-paper work to one component, repair it, reintegrate it and then verify again.

The Handbook describes a loop, not a staircase that can never be revisited.

Feedback is not complete until the learner uses it

Tutoring produces abundant feedback.

Because the tutor sits close to the learner, feedback can occur after almost every attempt.

That is both an advantage and a danger.

Too much feedback can turn the tutor into a continuous correction engine while the learner becomes a passive recipient.

The Feedback Receipt asks whether feedback changed the learner's next attempt.

The Feedback-Action Loop strengthens the same principle: feedback requires an opportunity for action before its value can be judged.

The Feedback Timing Gate asks what should be corrected now, what can wait and what the learner should first revisit independently.

The Correction Load Gate asks how many errors can be usefully addressed without turning a page into an unusable wall of commentary.

Feedback therefore has four parts:

1. Identify the gap relevant to the current target.
2. Provide information that the learner can use.
3. Give the learner an opportunity to act on it.
4. Inspect the next performance for change.

If part three is absent, the tutor has delivered commentary, not completed a feedback loop.

A useful tutoring habit is to reserve time for response to feedback, not merely time for giving it.

The best correction is not always immediate correction

Immediate correction can prevent an error from being practised.

Delayed correction can allow the tutor to see whether the learner notices the problem, can self-correct, or can complete a larger reasoning chain before interruption.

The correct timing depends on the job.

Correct immediately when an error is likely to contaminate every later step and the lesson is teaching a new procedure.

Delay when the tutor needs to inspect the learner's reasoning trajectory.

Ask the learner to review first when self-monitoring is part of the target.

Use a later comparison when seeing two versions side by side will make the difference clearer.

The Feedback Timing Gate prevents "correct everything immediately" from becoming doctrine.

The tutor's job is not to maximise the number of corrected mistakes.

It is to maximise the learning value of correction.

Reattempts should change what the tutor can infer

The Reattempt asks what a second attempt means.

If the learner retries the identical question immediately after seeing the answer, success may show recognition or short-term repair.

That can be useful during teaching.

It is weak evidence of durable independent learning.

A better reattempt sequence may move through conditions:

same question with explanation,

same question after the explanation is hidden,

minimally changed question,

fresh question after other work,

delayed return,

mixed context.

The conditions become gradually more demanding.

This is why the Handbook distinguishes repair from verification.

Repair gives the learner another chance with appropriate support.

Verification asks whether the target capability survives when the support condition changes.

Tutors should not use verification conditions too early and call the resulting failure proof that the teaching failed.

They should also not use supported repair conditions too late and call the resulting success mastery.

Transfer must be designed, not hoped for

A learner does not automatically generalise because the tutor explained a principle correctly.

Transfer becomes more likely when the tutor deliberately changes features while preserving the underlying structure.

The Example-Variation Gate asks how much variation is useful.

The Changed Question tests whether learning travels when the surface changes.

The Classroom–Tuition Translation Gate and Curriculum-Method Reconciliation Gate ask whether knowledge survives across different methods, notation and language between school and tuition.

The Modality-Shift Gate asks what changes when learning moves between in-person and live-online tuition.

The Response-Modality Equivalence Gate asks whether oral, typed, handwritten and visual responses actually measure the same target skill.

Transfer therefore includes more than harder questions.

It can mean:

new wording,

new representation,

new context,

new modality,

new time condition,

new tutor,

new method language,

new task ordering,

or a new combination of old components.

The tutor should choose the transfer condition that matters for the real performance.

Academic language should be made accessible without being removed

Tutors often simplify language to make a difficult idea easier.

That can be helpful.

It can also produce a second problem if the learner later meets the full academic language in school or examinations and cannot connect it to the simplified explanation.

The Explanation-Language Gate asks how to make an idea understandable without simplifying away the language the learner eventually needs.

A strong sequence may be:

plain-language explanation,

explicit mapping to the academic term,

worked examples using both forms,

learner explanation using the term,

fresh task in which the academic language appears without translation.

The goal is access followed by ownership.

Not permanent translation.

Alternative methods require validity before conformity

Learners often produce methods that differ from the tutor's preferred route.

A tutor may reject the method because it is unfamiliar, or accept it because it produced the right answer.

Both responses can be too shallow.

The Alternative Method Gate asks:

Is the method valid?

Is it general enough for the relevant class of problems?

Can the learner explain or reproduce it?

Is it efficient enough for the performance context?

Will the school or examination accept the representation?

Does it interfere with later learning?

A different method is not automatically a problem.

A method that works only accidentally is.

The tutor should judge the mathematical, scientific, linguistic or assessment validity of the route before deciding whether conformity is necessary.

The learner should eventually choose, not just execute

Many tutoring materials announce the method in advance.

A chapter title says "simultaneous equations".

A worksheet says "percentage increase".

A writing exercise says "use PEEL".

A comprehension page says "inference questions".

That is appropriate while a technique is being introduced.

It weakens later evidence because the hardest decision may be recognising which technique applies.

The Mixed Set, Question Return Gate, Challenge Readiness Gate and Worked-Example-to-Problem Alternation Gate all eventually move responsibility back to the learner.

The learner should encounter moments where the route is not announced.

That does not mean throwing novices into discovery.

It means removing labels after the relevant knowledge is stable enough for selection to become the next job.

Examination performance is an integration problem

A learner can possess many components and still underperform in a full paper.

The examination integrates:

knowledge,

question interpretation,

method selection,

timing,

attention,

error recovery,

writing quality,

stamina,

prioritisation,

and the ability to switch between task types.

The Performance Coach owns that broader role.

The Timed Set introduces the clock without confusing speed with learning.

The Full Paper observes the integrated system.

The Second Full Paper asks whether a targeted repair survives reintegration.

The Trend asks whether the result persists across more than one paper.

The Handbook therefore does not treat the examination as a final surprise.

Performance conditions are introduced deliberately after local capabilities exist.

The tutor can then locate whether failure is local—a component is weak—or systemic—the learner cannot coordinate several competent components under pressure.

Homework is evidence only after its provenance is understood

Homework can be useful practice.

It can also be a poor measure of independent learning.

Parents may help.

Friends may compare answers.

Model solutions may be visible.

AI may draft an explanation.

The learner may spend unlimited time.

The Support Provenance Check asks who or what helped produce the work.

The Homework Collision Gate asks when urgent school homework should replace the planned tuition route and when it should not.

The Submission Boundary protects assessed school work from becoming tutor-authored work.

The tutor should never punish legitimate help by pretending all homework must be unaided.

Instead, label the condition.

"Completed with parent explanation."

"Completed independently, checked with model answer after."

"AI used to generate examples; final response written without AI."

"Started alone, tutor helped after question three."

That information makes the work more interpretable.

It also teaches the learner that using support responsibly is different from hiding support.

A good tutor is trying to become less necessary for the target capability

The Handbook treats independence as a designed outcome.

This does not mean a tutor should disappear as soon as the learner improves.

A learner may become independent in one capability while still needing help elsewhere.

But every intervention should contain a direction of travel.

The Dose calibrates help.

The Fade reduces help.

The Independence Test verifies unsupported performance.

The Release asks when routine intervention can end.

The Check-In preserves light monitoring without rebuilding dependence.

The Re-Entry Gate controls when structured support should return.

The Re-Entry Plan designs temporary support with an exit from day one.

The Re-Entry Review decides whether support should continue, change, shrink or end.

The Stability Window holds the new route steady long enough to learn from it.

This sequence makes the ethical direction of tutoring explicit.

Support exists to build capability.

It should not quietly redefine success as continued consumption of support.

Independence is not "no help ever"

A competent learner uses resources.

They check notes.

They ask questions.

They compare methods.

They seek feedback.

They collaborate.

They use technology.

Independence means the learner can regulate when and how support is used without support carrying the target performance by default.

A mature learner may say:

"I tried two methods. I am stuck at this decision. Can you check whether I am interpreting the condition correctly?"

That is different from handing the problem to the tutor at the first sign of friction.

The Handbook therefore treats help-seeking as a capability, not a failure.

The tutor can teach the learner to attempt, locate the uncertainty, formulate the question and use the answer to resume control.

The goal is agency with evidence, not solitary struggle.

A three-student tutorial is not three one-to-one lessons compressed into one room

Small-group tuition creates a different teaching system.

The tutor has several learners present at the same time. They can hear one another's explanations, compare approaches, observe errors, wait while another learner receives support, and sometimes solve together.

That creates opportunities unavailable in one-to-one tuition.

It also creates failure modes unavailable in one-to-one tuition.

One learner can answer first and destroy the next learner's independent evidence.

One confident learner can become the implicit pace-setter.

One learner can appear to understand because the group produced a correct answer.

The tutor can divide time equally and still distribute learning opportunity badly.

A shared explanation can be efficient when all three learners need the same repair.

The same shared explanation can be wasteful when the apparent common error actually has three different causes.

The Handbook therefore treats small-group tutoring as orchestration, not merely a ratio.

The Education Endowment Foundation's current small-group tuition summary defines small-group tuition as a teacher, trained teaching assistant or tutor working with two to five pupils, and reports that impact varies with targeting, teaching quality, subject, group composition and implementation. The useful lesson for this Handbook is not to turn an average into a promise. It is that group size creates an opportunity for more intensive interaction, but the educational value depends on what the tutor does with that opportunity.

The Handbook's three-learner articles operationalise that problem.

Keep individual evidence inside shared learning

The Three-Learner Orchestration article asks how questioning, attempt time, feedback and observation should be distributed without mechanically equalising everything.

The Individual Accountability Gate asks how to use collaboration without allowing the group's best answer to stand in for each learner's understanding.

The Peer Answer Leakage Boundary protects independent evidence before discussion.

The Peer-Explanation Boundary allows one learner to explain without converting that learner into the unpaid assistant teacher or allowing the others to borrow understanding indefinitely.

The Shared-Error Reteach Gate asks when a common-looking mistake justifies one collective repair and when the group should branch.

The Grouping Gate and Regrouping Stability Gate ask who should learn together and when an existing group should remain, branch temporarily or be reorganised.

The Ratio–Tutor-Skill Fit Gate asks whether three learners can actually be taught as one tutorial given the tutor's skill and the diagnostic intensity of the cases.

The Tutor Case-Mix Allocation Gate extends this to programme staffing: equal headcount is not necessarily equal teaching load.

These are separate jobs because a group can be appropriately formed and still badly orchestrated, or well orchestrated and still contain a learner whose needs have diverged too far.

Equal treatment is not identical treatment

A three-student tutorial should be fair.

Fairness does not require identical teacher minutes, identical numbers of questions or identical feedback at every moment.

One learner may need a brief explanation.

Another may need uninterrupted attempt time.

A third may need a fresh verification item.

If the tutor gives each learner five minutes merely because five minutes feels equal, the distribution can be educationally unequal.

A more useful fairness principle is:

each learner receives the learning opportunity and evidence condition required for the current job, while no learner is systematically deprived of tutor attention, challenge, participation or independent accountability.

The tutor should be able to explain the asymmetry.

"Student A is working independently because the last two tasks show the method is stable; I am using this time to diagnose Student B's different error. Student C will receive a fresh item before we regroup."

That is not neglect.

It is controlled differentiation.

Protect the quiet learner from invisibility

Small groups naturally reward visible participation.

The fast responder speaks.

The confident learner volunteers.

The articulate learner explains.

The quieter learner can disappear inside apparent group success.

The Handbook counters this with private evidence moments:

• silent first attempts,
• mini-whiteboard answers,
• written predictions,
• individual method choices before discussion,
• private explanations,
• staggered questioning,
• and fresh items after peer talk.

The tutor does not need to turn every lesson into continuous testing.

A few strategically placed individual receipts are enough to reveal whether the group is carrying someone.

This is where Evidence Sample, Individual Accountability, Peer Answer Leakage and Learner-Talk Interpretation meet.

More talk is not automatically more thinking.

Less talk is not automatically less learning.

The tutor needs evidence that participation is connected to the target cognition.

Group composition should be temporary enough to respond to evidence

Grouping by current need can be useful.

Grouping by permanent ability identity can become harmful.

The Grouping Gate asks whether learners share enough of the same instructional route to benefit from learning together.

The Regrouping Stability Gate asks when the group should remain stable and when divergence is consequential enough to justify a change.

The stability side matters because constant regrouping has costs.

Relationships reset.

Routines reset.

The tutor's accumulated knowledge of interaction patterns becomes less useful.

The learners may interpret every movement as ranking.

The regrouping side also matters because a group can become educationally incoherent if one learner is consistently several decisions away from the others.

The gate is not "Are their marks close?"

It is "Can one tutor orchestrate a coherent lesson in which each learner has a productive route and visible evidence?"

Peer explanation should be used as learning, not labour substitution

Asking a learner to explain to a peer can deepen learning.

It can also become a hidden staffing strategy in which the strongest learner teaches while the tutor attends elsewhere.

The Peer-Explanation Boundary protects against that drift.

Peer explanation is most useful when:

• the explaining learner benefits from articulating reasoning,
• the listening learner has enough foundation to evaluate or question,
• the tutor remains responsible for accuracy,
• and the role does not become fixed.

The explaining learner should not become "the smart one".

The listening learner should not become "the weak one".

Roles should move as the job changes.

One learner may be strongest in algebraic structure, another in checking units, another in constructing precise English explanations.

The group should reveal capabilities, not freeze identities.

One group's apparent success can hide three different routes

Suppose three learners all answer the same Science question incorrectly.

The tutor sees a shared error and prepares one reteach.

Before teaching, the tutor asks each learner to explain the reasoning.

Learner A misunderstood the scientific process.

Learner B understood the process but misread the diagram.

Learner C understood both but copied A's answer after discussion.

The written error was shared.

The teaching job was not.

That is why the Shared-Error Reteach Gate exists.

A group response is a starting signal.

The tutor should verify whether the mechanism is shared before committing everyone to the same repair.

The reverse also occurs.

Three different wrong answers can arise from one common misconception.

Surface variation does not prove different causes.

The group tutor must diagnose at the level that changes the next action.

Trust matters because teaching requires exposure

A learner must reveal uncertainty for tutoring to work well.

They must attempt before they are certain.

They must show incomplete reasoning.

They must sometimes admit that they do not understand an explanation that seemed clear to everyone else.

A relationship with sufficient trust makes that possible.

The Relationship Quality Check asks whether rapport is supporting learning without treating liking the tutor as evidence of learning.

The Relationship Repair asks how to rebuild working trust after friction without lowering the academic standard.

The Professional Relationship Boundary protects warmth from expanding into secrecy, inappropriate personal contact or roles outside tuition.

The Handbook's position is deliberately balanced.

Cold technical instruction can suppress useful learner evidence.

Warmth without academic clarity can produce comfortable stagnation.

A tutoring relationship should make truthful learning evidence easier to produce.

Productive safety is not the removal of challenge

A learner can feel respected while still being asked to attempt difficult work.

They can trust the tutor while receiving correction.

They can dislike a task and still benefit from it.

The tutor should not confuse immediate comfort with educational value.

At the same time, challenge should not be used as an excuse for humiliation, pressure or avoidable confusion.

The Challenge Readiness Gate asks whether harder work extends learning or simply hides unstable foundations.

The Learner Voice Interpretation Gate asks how student feedback about difficulty, helpfulness and fit should inform decisions without treating satisfaction as proof of learning.

The Implementation Acceptability Gate later applies a related principle at programme level: reactions to a new practice matter because unacceptable practices are difficult to sustain, but popularity is not the same as evidence of effectiveness.

Motivation is often a description after the fact

When learners do not engage, adults often say they lack motivation.

That label can end inquiry.

The Handbook instead opens a differential.

Is the task understandable?

Does the learner know how to start?

Is the work far below or above the current capability?

Is there a relationship problem?

Is the learner tired?

Was tuition chosen by the learner?

Is attendance disrupting continuity?

Is the learner protecting themselves from repeated failure?

Is the group condition uncomfortable?

Does the learner see the purpose?

The Enrolment–Engagement Boundary is especially important for learners who did not choose tuition. Attendance does not prove motivation. Compliance does not prove learning. Resistance does not automatically prove defiance.

The Attendance Differential separates logistics, load, relationship and avoidance.

The Behaviour–Learning Differential separates behavioural symptoms from instructional causes.

The tutor can still set expectations.

But expectations should sit on a correct diagnosis.

Learner voice is evidence, not command and not noise

Learners know things the tutor cannot directly observe.

They know which explanations feel confusing.

They know when they are guessing.

They know whether homework support was used.

They know which tasks create overload.

They know whether a tutor's pace feels impossible to follow.

Their reports matter.

They are also fallible.

A learner may prefer easy practice that produces confidence.

They may dislike a valuable difficult activity.

They may misidentify the cause of a problem.

The Learner Voice Interpretation Gate therefore treats student feedback as one evidence source.

The tutor listens, checks against performance evidence, and uses the feedback to improve the route.

That is respect without abdication.

Family observations can expand the evidence window

Parents and caregivers see learning outside tuition.

They may notice that homework now takes half the time.

They may see avoidance that never appears in a lesson.

They may report that the learner can explain a concept at home.

They may also see only a slice of the process, interpret a single event strongly, or compare the child to siblings.

The Family-Evidence Integration Gate asks how family observations can inform the route without replacing direct learning evidence.

The useful practice is to convert broad reports into inspectable signals.

"He is more confident" can become:

"He now starts homework without asking for the answer key first."

"She is careless" can become:

"She completed the last three pieces quickly and did not reread instructions."

"He understands at home" can become:

"He explained the method to me without notes on two evenings."

These observations can guide the next diagnostic check.

They should not become permanent labels.

Parent communication should preserve the child as a learner, not a score

The Parent Review established an early Handbook principle: explain progress, uncertainty and the next decision without turning the child into a score.

A useful parent review answers:

What are we currently trying to improve?

What evidence has changed?

What remains uncertain?

What is the next decision?

What can home support do without duplicating tuition?

What are we not claiming yet?

This is more useful than a report consisting of worksheet percentages.

It also reduces pressure to perform certainty.

Parents do not need the tutor to pretend every week has a clear upward line.

They need an intelligible route.

Priority conflicts should be made explicit

Learner, family and school may want different things.

The learner wants help with tomorrow's homework.

The parent wants examination gains.

The school needs completion of a project.

The tutor sees a prerequisite weakness that controls all three.

The Priority Arbitration Gate exists because quietly choosing one voice creates hidden conflict.

The tutor should state the trade-off.

"If we spend the whole lesson finishing the project, we can meet the deadline but we will postpone the prerequisite repair. We can use twenty minutes to unblock the project, then protect the remaining time for the underlying skill."

That is professional coordination.

It does not guarantee everyone gets everything.

It makes the decision visible.

Tuition should not become unofficial case management

As relationships deepen, tutors may be asked to coordinate school, family, specialists, schedules, behaviour plans and emotional support.

Some coordination is necessary.

The Support-Network Coordination Gate asks how to keep school, family and tuition coherent without creating parallel plans, duplicate work or an unofficial case-management role.

The Scope Boundary states the wider rule.

Tutors should teach, observe, communicate relevant learning evidence and coordinate where educationally necessary.

They should not assume roles that require different professional authority or expertise.

The boundary protects the learner and the tutor.

Cultural responsiveness requires adaptation without stereotyping

The Cultural-Safety Adaptation Gate asks how language, examples and participation conditions can be adapted without turning culture into a diagnosis.

A learner's cultural or linguistic background can matter.

It should not become an explanation for every difficulty.

The tutor can choose examples that do not assume unfamiliar experiences, pronounce names correctly, avoid humiliating participation structures, clarify idiom, and invite the learner to signal when examples do not make sense.

At the same time, academic expectations remain.

Adaptation should increase access to the target, not lower the target because of an identity assumption.

Tutors should be classified by the job they can perform, not by prestige labels

The Handbook begins with tutor roles because "good tutor" is too vague to support matching.

A learner may need:

a Homework Helper,

an Explainer,

a Drill Builder,

a Diagnostic Tutor,

a Route Designer,

a Performance Coach,

or a Learning Architect.

The same person may perform several roles.

The point is functional fit.

A tutor who is excellent at explaining may not be strong at diagnosis.

A tutor who creates disciplined practice may not yet be skilled at building a long-term route across school, tuition and examination performance.

A highly credentialed subject expert may still need training in questioning, feedback, small-group orchestration or interpreting learner evidence.

A charismatic tutor may produce strong engagement while unintentionally carrying too much of the thinking.

The Tutor–Student Fit, Tutor–Learner Matching Gate, Tutor Selection Gate, Tutor Selection Evidence Gate, Tutor Capacity Boundary and Ratio–Tutor-Skill Fit Gate progressively deepen the staffing problem.

Selection should look for observable teaching evidence

Credentials can matter.

Subject knowledge matters.

Experience can matter.

But the Tutor Selection Evidence Gate asks a more operational question: what observable evidence shows that the tutor can perform the teaching jobs required?

A selection process might inspect:

• accuracy of subject explanations,
• ability to diagnose from learner work,
• quality of questions,
• response to a partly correct answer,
• capacity to adapt without abandoning the learning goal,
• awareness of professional boundaries,
• handling of uncertainty,
• ability to use evidence rather than charisma,
• and reflection after teaching.

One polished demonstration lesson should not carry the whole decision.

A demo can be rehearsed.

An interview can reward fluency.

Credentials can be proxies.

The selection system should combine evidence sources whose weaknesses differ.

Capacity is an instructional variable

A tutor can be skilled and overloaded.

Preparation deteriorates.

Records become thin.

Follow-up disappears.

Feedback becomes generic.

The tutor relies on familiar worksheets because there is no time to diagnose.

The Tutor Capacity Boundary treats workload as part of teaching quality.

Programme leaders should therefore ask not only "How many learners can this tutor take?"

They should ask:

How many diagnostically intensive cases?

How many different subjects or levels?

How much preparation?

How much follow-up?

How many group configurations?

What coaching responsibilities?

What record burden?

The Tutor Case-Mix Allocation Gate later sharpens this: equal headcount is not equal work.

Staffing should account for case complexity.

Fit should be updateable

A tutor–learner match can work at one stage and become less suitable later.

A learner may move from explanation needs to high-level performance coaching.

A tutor may be excellent for foundational repair but not for a specialised curriculum component.

The relationship may deteriorate.

Schedule constraints may change.

The Tutor Reassignment Gate allows fit to be reconsidered without treating reassignment as failure and without resetting the learner.

The Continuity Packet, Receiving Check and First Change protect accumulated learning history when a new tutor takes over.

Continuity means the new tutor inherits evidence, not conclusions that can never be questioned.

The receiving tutor should verify the route before changing it.

That balances respect for prior work with fresh professional judgement.

Training is not implementation

A tutor can attend a workshop, describe a method accurately, pass a quiz and still not use the method well in a real lesson.

That gap explains a major expansion in the later Tutor Handbook volumes.

The series moves beyond "What should a tutor do?" into "How does a tuition programme know tutors can do it reliably under ordinary conditions?"

The Training Need Gate asks whether a problem is actually a tutor skill gap or instead a materials, workload, curriculum or system problem.

The Rehearsal-to-Live Gate asks when professional judgement practised in cases or simulations is ready for live use.

The Coaching Focus Gate asks how to choose one high-leverage teaching move rather than producing a long observation checklist.

The Coaching Receipt asks whether professional feedback changed live teaching.

The Training-to-Live-Practice Transfer Check asks whether a tutor can use a trained move with a real learner rather than merely describe it.

The Coaching-Support Taper Gate asks when observation and coaching can reduce.

The Professional-Learning Cycle Gate integrates planning, modelling, rehearsal, live coaching and review.

Together they establish a strong principle:

professional learning is complete only when the intended change appears reliably in ordinary teaching and can be sustained with an appropriate level of support.

Observe teaching without turning it into theatre

Observation changes behaviour.

A tutor who knows a coach is present may prepare more carefully, choose a safer lesson, perform routines unusually explicitly or avoid difficult cases.

That does not make observation useless.

It makes sampling important.

The Observation Window asks how much live teaching evidence a coach needs before giving feedback.

The Observation Reactivity Check asks how the presence of a parent, coach or camera may change the lesson.

The Observation-Sample Gate asks which sessions a programme should review.

The Tutor-Observation Sample Gate asks how a coach should choose lessons and frequency without mistaking a staged session for normal practice or turning coaching into surveillance.

A defensible observation system should therefore sample across:

• ordinary and difficult lessons,
• different learner profiles,
• different points in a topic,
• planned and contingent moments,
• and enough time for the target teaching move to appear naturally.

The purpose is not to catch tutors failing.

The purpose is to build an accurate model of practice.

Surveillance changes the relationship.

Coaching should create professional evidence that tutors can use, not a permanent threat of inspection.

Coaching feedback should have a receipt

A coach says:

"Ask more diagnostic questions."

The tutor agrees.

Nothing has changed yet.

A better coaching loop defines an observable move.

"Before you explain a wrong answer, ask one question that separates missing knowledge from misread task conditions."

The tutor rehearses the move.

The coach observes a later lesson.

The tutor uses the question.

The learner response changes the chosen intervention.

Now there is a receipt.

The Coaching Receipt mirrors the learner Feedback Receipt.

Agreement is not implementation at either level.

A professional learning system should ask what changed in behaviour and what changed because of that behaviour.

Consistency should protect active ingredients, not create scripts

A tuition programme wants tutors to teach coherently.

Families should not receive entirely different philosophies depending on which tutor happens to be assigned.

Learners should not have to relearn programme routines every time staffing changes.

But consistency can become scripting.

The Instructional-Practice Consistency Gate asks how a programme keeps the same teaching model coherent without turning good teaching into rigid performance.

A useful consistency model has layers.

Non-negotiable principles might include:

• diagnose before large intervention,
• verify learning on fresh evidence,
• preserve professional boundaries,
• do not submit tutor-authored work as learner work,
• verify AI-generated materials before use,
• record support conditions when they change interpretation,
• and keep a route towards learner independence.

Flexible implementation might include:

• the exact wording of a question,
• examples chosen,
• pacing,
• visual representation,
• order of two equivalent practice tasks,
• or how a learner explains reasoning.

The programme should standardise the job and quality boundary more strongly than the surface form.

Professional disagreement should be resolved with evidence, not hierarchy

Two tutors can look at the same learner work and disagree.

A coach may disagree with a tutor's interpretation of a lesson.

The programme lead may favour a method that the classroom evidence does not support.

The Evidence Conference and Coaching Disagreement Protocol make disagreement legitimate and structured.

A useful protocol asks:

What decision are we trying to make?

What evidence is shared?

Where does interpretation diverge?

What additional evidence would separate the views?

What action is reversible while uncertainty remains?

When will we review?

This is better than voting.

It is better than averaging incompatible interpretations.

It is better than pulling rank.

Professional authority still exists. Some decisions belong to programme leaders or safeguarding procedures.

That is why the Escalation-Authority Gate distinguishes decisions a tutor can make alone from those that must move elsewhere.

An evidence-informed practice can fail because the implementation failed

A programme adopts a promising instructional routine.

Tutors attend training.

The materials are distributed.

Three months later, learner outcomes are unchanged.

What does that mean?

Possibilities include:

• the practice was ineffective in this context,
• tutors did not use the practice as intended,
• the practice was too burdensome for ordinary workload,
• only some tutors used it,
• only easy learner cases received it,
• the active ingredient was removed during adaptation,
• training did not transfer,
• materials were incoherent,
• monitoring changed behaviour only temporarily,
• or the programme stopped supporting the practice after launch.

The later Tutor Handbook volumes progressively separate those explanations.

The Implementation Fidelity Check asks whether the route failed or was never actually run as intended.

The Active-Ingredient Adaptation Gate asks whether local changes preserved what makes the routine work.

The Implementation Feasibility Gate asks whether the routine fits ordinary tutor time, materials and learner variation.

The Implementation Reach Gate asks whether the practice reached the learners and tutors it was meant to serve.

The Implementation Sustainability Gate asks whether the practice can survive after launch without one champion or permanent extra effort.

The Implementation Acceptability Gate asks how tutor, learner and family reactions should inform the rollout without equating popularity with effectiveness.

The Implementation Readiness Gate asks whether the programme is actually ready to launch.

This architecture prevents the phrase "implementation problem" from becoming a vague excuse.

Each implementation dimension becomes inspectable.

Readiness comes before rollout

The Implementation Readiness Gate sits logically before training everyone.

A programme should ask:

Is the problem clear enough to justify the practice?

Is there leadership ownership?

Do tutors understand why the change is needed?

Are materials ready?

Is there time for training, rehearsal and coaching?

Can current records capture the necessary evidence without excessive burden?

Are there enough coaches or programme leads to support the change?

What must remain stable during the trial?

What is the rollback plan if the change creates harm?

How will learner and tutor feedback be collected?

What outcome would justify expansion?

A weak readiness state creates predictable failure.

If the timetable cannot accommodate the routine, fidelity problems are not evidence of lazy tutors.

If materials are contradictory, more training may not solve the problem.

If there is no coaching capacity, a workshop may produce awareness without practice change.

The gate puts responsibility on the system before blaming individuals.

Feasibility is not the same as effectiveness

A method can work under ideal conditions and be infeasible at scale.

Suppose a new feedback routine requires fifteen minutes of individual written commentary per learner after every session.

The feedback may be excellent.

A tutor with several groups may not be able to sustain it.

The programme could respond by demanding harder work.

The Implementation Feasibility Gate asks whether the routine can fit ordinary staffing, time, materials and learner variation.

Feasibility does not prove educational value.

It asks whether the system can realistically deliver the practice well enough for its educational value to matter.

A practice that only works when everyone performs unpaid extra labour has not yet been implemented sustainably.

Reach asks who is missing

A programme may report successful implementation because the tutors who adopted the practice show strong results.

What about the tutors who did not adopt it?

What about learners with more complex needs?

What about online sessions?

What about less experienced tutors?

What about groups that were excluded because the practice was difficult to organise?

The Implementation Reach Gate prevents easy cases from defining success.

Reach should be examined before celebrating programme-wide implementation.

A practice that succeeds only for the people easiest to serve may still be useful, but the claim must match the actual reach.

Acceptability matters without becoming a popularity contest

Tutors can resist a useful practice because it is unfamiliar.

They can also resist because it is poorly designed, too burdensome, conflicts with curriculum reality or produces unintended consequences.

Learners can dislike productive difficulty.

They can also detect confusing or humiliating implementation earlier than leaders do.

Families can prefer familiar routines even when a new approach is stronger.

They can also identify practical burdens the programme missed.

The Implementation Acceptability Gate treats reactions as evidence about implementation conditions.

It asks:

What exactly is being resisted?

By whom?

At what point?

Is the concern about the learning mechanism, the workload, the communication, the schedule, the materials or the values of the practice?

What can be adapted without removing the active ingredient?

Acceptability matters because people must actually use the practice.

Popularity is not proof that the practice works.

Sustainability asks whether the practice survives normal life

Launch periods are unusual.

Leaders pay attention.

Tutors receive reminders.

Extra resources appear.

Coaches observe frequently.

Everyone knows the initiative matters.

The Implementation Sustainability Gate asks what happens later.

Can the routine survive tutor turnover?

Can new staff learn it?

Are materials maintained?

Can coaching taper without practice collapsing?

Does the practice still occur during busy examination periods?

Does the programme have a simple way to detect drift?

Can the practice be improved without becoming unrecognisable?

Sustainability is the difference between an initiative and an operating system.

Professional-learning cycles connect implementation to tutor practice

A workshop changes knowledge.

A professional-learning cycle can change practice.

The Professional-Learning Cycle Gate connects:

problem definition,

knowledge building,

modelling,

rehearsal,

live use,

observation,

feedback,

revision,

and later review.

AERO's current implementation resources similarly place professional learning inside deliberate implementation rather than treating a workshop as the endpoint.

The Handbook's contribution is to translate that principle into tutor-programme decisions.

A programme should know which tutor behaviour it is trying to improve, what competent use looks like, how tutors will rehearse, where live evidence will come from and how support will taper after reliable use appears.

Record enough to preserve continuity, not enough to create a second job

Tutoring generates data easily.

Scores.

Annotations.

Session notes.

Homework.

Screenshots.

AI transcripts.

Parent messages.

Audio or video.

Observation forms.

Timing data.

Progress charts.

The risk is to assume more record equals more professionalism.

The Record Minimum asks what evidence is necessary for continuity and improvement without collecting more personal data than the educational job requires.

The Recording Window asks when a tutor should write notes without missing the live reasoning that needs observation.

The Evidence-Capture Burden Gate asks when recording begins to compete with teaching.

The Learner-Data Retention Gate asks when evidence should be kept, anonymised, shared or deleted after its instructional purpose expires.

The Learner-Progress Data Ownership Gate asks how learners can see and use their own progress evidence.

Together these articles establish data minimisation as an instructional principle, not merely an administrative one.

If recording reduces the tutor's ability to observe the learner, the system may be collecting evidence while damaging the event it wants to understand.

A useful record preserves decisions and conditions

The most valuable record is often not a complete transcript.

It is a compact decision record.

Target capability.

Observed evidence.

Support condition.

Current hypothesis.

Action taken.

Receipt expected.

Review point.

That is enough to preserve why the route changed.

The Decision Record exists because hindsight rewrites stories.

After improvement, the old problem can look obvious.

After failure, the original decision can look foolish.

A contemporaneous record protects learning from the programme's own memory bias.

Learners should see evidence without becoming the evidence

Showing progress can support goal setting and agency.

It can also turn a score into identity.

The Learner-Progress Data Ownership Gate asks how to make evidence available to the learner while preserving interpretation.

Instead of:

"You are a 62% student."

Use:

"Across the last three fresh sets, accuracy in method selection improved from 4 of 10 to 7 of 10. Calculation accuracy is already stronger. Our next job is choosing the method before the clock is added."

The learner receives useful information.

The evidence remains a description of performance under conditions.

It does not become a permanent category of person.

AI does not inherit professional judgement merely because it produces fluent text

Generative AI can create questions, explanations, examples, summaries, feedback and practice variations quickly.

That capability can increase tutor productivity.

It can also manufacture plausible errors at scale.

The AI Material Verification Gate sets the core rule: unverified generated material should not reach the learner merely because it sounds polished.

Verification includes:

• factual and subject accuracy,
• curriculum fit,
• answer correctness,
• ambiguity,
• difficulty,
• accessibility,
• whether examples accidentally reveal the answer,
• whether distractors test the intended misconception,
• and whether the material duplicates private or protected content inappropriately.

The tutor remains accountable for what is taught.

AI support changes the provenance of learner work

A learner can use AI to:

brainstorm,

check grammar,

explain a concept,

generate examples,

solve a problem,

draft an answer,

or evaluate a response.

These uses are not equivalent.

The Support Provenance Check and Human-Supported AI Engagement Gate ask what the tool actually carried.

Screen time is not learning opportunity by itself.

A learner who asks AI for the final answer and copies it has produced little independent evidence.

A learner who attempts first, asks for a hint, compares the hint with their reasoning and then completes a fresh problem may use AI productively.

The tutor should design the help condition.

"Ask for one hint, not a full solution."

"Use AI to generate three practice questions, then solve them without AI."

"Ask for criticism of your paragraph, but decide which changes to accept and explain why."

"Use AI to explain the concept twice, then close it and teach the concept back."

The tool should sit inside a learning loop.

AI can increase material volume faster than curriculum coherence

A tutor can now generate fifty questions in seconds.

That does not mean fifty questions belong in the route.

The Instructional Material Coherence Gate becomes more important when generation is cheap.

Practice should still fit the sequence, difficulty, representation, target capability and evidence purpose.

An AI-generated question bank can create the illusion of personalisation while quietly fragmenting the curriculum.

Personalisation should mean better fit to diagnosed need.

Not endless variation.

Human support remains necessary where judgement is consequential

AI can propose a diagnosis.

It cannot be allowed to convert uncertain learner evidence into an identity.

AI can generate feedback.

A tutor must decide whether the feedback targets the right learning job and whether the learner can use it.

AI can estimate difficulty.

The tutor should inspect actual learner performance.

AI can summarise progress.

The tutor must preserve the conditions and uncertainty behind the summary.

The Human-Supported AI Engagement Gate therefore asks whether the human presence is producing genuine learning opportunity or simply supervising technology use.

The tutor's value is not that a machine cannot generate words.

The tutor's value is responsibility for interpretation, sequencing, relationship, evidence, boundaries and consequential decisions.

Tuition should not force the learner to run two incompatible curricula

A learner attends school and tuition.

The school uses one method, notation or sequence.

The tutor prefers another.

Sometimes the tutor's method is clearer.

Sometimes the school method is required for assessment.

Sometimes both are valid.

The learner can end up maintaining two parallel systems and switching under pressure.

The Curriculum Translation Check, Classroom–Tuition Translation Gate and Curriculum-Method Reconciliation Gate address this problem at increasing depth.

The tutor should ask:

What does the school require?

Which differences are superficial?

Which differences change the underlying concept?

Can the two methods be mapped explicitly?

Which method should the learner use under examination conditions?

Will teaching an alternative now reduce or increase confusion?

A good tutor translates when possible before replacing.

Materials should support one route

The learner may bring:

school notes,

tuition notes,

assessment books,

past papers,

teacher worksheets,

AI practice,

online quizzes,

model answers,

and parent-purchased resources.

More resources can create less coherence.

The Instructional Material Coherence Gate asks whether materials serve the current route or merely accumulate.

A resource earns a place when it performs a job:

teaches,

provides guided practice,

tests transfer,

samples progress,

or verifies performance.

If the tutor cannot state the job, the resource may be clutter.

Learning objectives need translation into visible success

Schools often express broad objectives.

"Write effectively for purpose and audience."

"Understand fractions."

"Analyse language."

"Apply scientific concepts."

The Learning-Objective Translation Gate asks how the tutor turns a broad goal into learner-visible success criteria without shrinking the capability into a mechanical checklist.

Useful success criteria identify what competent performance contains.

They should not reveal the exact answer.

A learner should know what they are trying to produce while still having to think.

This balances transparency and cognition.

Time spent on one improvement displaces another opportunity

The Learning Budget makes a neglected fact explicit: a tutoring hour is finite.

If twenty minutes are spent fixing presentation, those minutes are not spent on reasoning.

If an additional weekly session is added, it may displace rest, sport, family time, schoolwork or independent practice.

The Schedule-Substitution Gate asks whether another lesson is worth what the extra tuition displaces.

This is an educational question, not merely a financial one.

More tuition can help.

More tuition can also produce fatigue, dependency or reduced time for consolidation.

The decision should consider opportunity cost.

Helpful interventions can create new problems

The Intervention Cost asks whether a useful fix creates a side effect.

A mnemonic may improve recall but reduce understanding if it becomes the only route.

A highly structured writing frame may improve organisation but produce dependency.

Extra checking may reduce careless errors but destroy timing.

Intensive tutor prompting may improve lesson performance but reduce independent initiation.

A group change may improve fit but disrupt trust.

A new AI tool may save planning time but introduce verification burden.

The tutor should inspect both the target gain and the cost created by the intervention.

More sessions are not always the answer

The Dosage Differential asks whether the learner needs more tutoring or different tutoring.

If the current route is wrong, increasing dosage scales the wrong intervention.

If the route is sound but too infrequent for practice and feedback cycles to hold, additional dosage may help.

This distinction prevents volume from substituting for diagnosis.

The EEF's one-to-one tuition and small-group tuition summaries both emphasise targeted support, links to normal learning and staff quality. The Handbook uses that evidence as context, not as a fixed prescription for private tuition frequency.

The exact dose should follow the learner's route, the opportunity cost and the evidence that the present cadence is or is not producing change.

Articles should hand the reader to the next job

A strong Handbook article does not try to solve every adjacent problem.

It solves its job and then names the likely next jobs.

That creates a decision network.

Consider the pathway around a new learner.

First Lesson asks what the tutor should learn before teaching too much.

Second Lesson tests the first hypothesis.

First Month uses repeated lessons to stabilise the teaching route.

First Review asks whether to continue, advance, reclassify or stop.

Parent Review and Student Review bring different voices into the decision without allowing either voice to replace direct learning evidence.

Learning Contract clarifies responsibility.

Handover ends each lesson with a next move the learner can carry.

No article needs to reproduce the others.

The route exists because each article hands off cleanly.

A later pathway might look like:

Diagnostic Probe → Branch Point → Dose → Fade → Independence Test → Release → Check-In → Re-Entry Gate

Another:

Trial Run → Adoption Gate → Drift Check → Temporary Override → Rollback → Recovery Check → Rebuild → Cutover → Decommission

Another:

Training Need Gate → Coaching Focus Gate → Rehearsal-to-Live Gate → Coaching Receipt → Training-to-Live-Practice Transfer Check → Coaching-Support Taper Gate → Professional-Learning Cycle Gate

Another:

Implementation Readiness → Feasibility → Reach → Acceptability → Sustainability

The root manual should help the reader recognise these routes without forcing every future article into one fixed numbering taxonomy.

The number is an address.

The job is the meaning.

The series can grow without losing control

A long-lived handbook should expect new decisions to appear.

New technology changes evidence conditions.

New curriculum forms create translation problems.

New programme structures create implementation problems.

Research may sharpen a boundary.

A recurring failure mode may deserve its own owner.

The correct response is not to stop the series at an arbitrary number.

The correct response is to require a job test before adding a volume.

A proposed new article should pass five questions.

Is there a recurring decision?

A topic is not enough.

"AI in tuition" is a topic.

"How does a tutor verify AI-generated instructional material before it reaches a learner?" is a decision.

Is the decision consequential?

Would a wrong choice meaningfully affect learning, evidence, safety, workload, fairness, continuity or professional quality?

Is the decision already owned?

Search the existing Handbook by mechanism, not just title words.

"Observation sample" and "tutor observation sample" may look different but can collide unless their levels are separated.

One may own programme sampling across tutors.

Another may own a coach's sampling within a tutor's practice.

The boundary must be explicit.

Can the article state a receipt?

If the article cannot say what successful use changes, the job may be too vague.

Can the article hand off?

A useful owner sits inside the network.

It should know what comes before and after.

If a proposed article merely restates a broad philosophy with no distinct hand-off, it belongs in the root manual or another owner rather than as a new volume.

Duplicate volume numbers are a registry problem, not a reason to erase useful articles

The live Tutor Handbook contains early numbering collisions.

Examples around the 0135–0141 range include parallel published owners sharing a volume number.

There are also earlier parallel entries such as two published versions of the Change Queue under Vol. 0064.

The correct repair is not to pretend those pages never existed.

They are public, linked and may own useful distinctions.

The root registry should preserve the published URLs while making the job difference visible.

Future publication should avoid creating new collisions by checking the live registry before assigning a number.

Where two existing pages are genuinely duplicates in job, a later editorial pass can decide whether one becomes a redirect, appendix, companion or retained historical page. That is a separate canonical-repair job and should not be mixed into this operating-manual rebuild.

For now, registry integrity means:

• preserve public URLs,
• show both published owners where they exist,
• distinguish their jobs when possible,
• flag the numbering collision internally,
• and never reuse the number again.

Vol. 0027 remains an explicit gap

A missing number is less harmful than an invented article.

The live series moves from 0026 to 0028.

The root should therefore continue to state that Vol. 0027 is not currently available publicly.

If a future decision deserves that address, the editorial team may choose whether to fill the gap or leave it as a historical registry state. The number itself should not pressure the series into manufacturing content.

Content follows the job.

The sequence does not manufacture a job merely to make the numbers look tidy.

Case 1: The learner who looks successful in tuition and underperforms at school

This composite case is fictional.

Nadia attends weekly Mathematics tuition in a three-student group.

In tuition, she usually scores above 80% on worksheets. Her school tests remain near 60%.

A weak tutoring system chooses one of two stories immediately.

"School tests are harder."

or

"Nadia is careless under pressure."

The Handbook route starts by defining the discrepancy.

The tutor compares the tasks.

The tuition worksheets are grouped by topic. The school paper mixes several topics and does not name the method.

That activates the Task-Purpose Gate, Task Purity Check and Mixed Set.

The tutor gives Nadia a fresh mixed set under untimed conditions.

Her calculation accuracy remains strong, but method selection drops.

The tutor now has a better hypothesis: the tuition route has been building execution more strongly than discrimination.

The Diagnostic Probe uses several items with plausible competing methods.

Nadia can perform each method when named. She struggles to identify the structural cue that selects the method.

The tutor does not add more blocked worksheets.

The Example-Variation Gate and Contrast-Case Gate are used to place similar-looking problems side by side and identify the decisive difference.

The Worked-Example-to-Problem Alternation Gate moves Nadia from studying annotated decisions to solving paired problems.

The Mixed Set then removes labels.

After several sessions, the tutor uses the Fresh-Item Reserve for a clean check.

Performance improves.

Only then does the Timed Set introduce time pressure.

The Full Paper later asks whether selection holds inside whole-paper performance.

The Second Full Paper checks whether the repair survived reintegration.

The Trend waits for enough evidence before declaring stable improvement.

Notice what did not happen.

The root manual did not say "do mixed practice" and end the case.

Each owner handled one decision.

The route changed because evidence changed.

Case 2: The learner who understands with the tutor and freezes alone

This composite case is fictional.

Marcus can follow algebra examples in tuition.

When the tutor asks questions during the demonstration, he answers correctly.

At home, he often leaves similar problems blank.

The tutor first checks Support Provenance.

During tuition, the tutor is supplying several small prompts:

"Which variable should we eliminate?"

"What could you multiply by?"

"Should the equations be added or subtracted?"

Marcus is doing calculations, but the tutor is carrying method initiation.

The Correct-Answer Audit confirms that successful tuition answers do not yet represent full route ownership.

The Dose asks which support can be reduced without causing total collapse.

The Worked-Example-to-Problem Alternation Gate changes the sequence.

Marcus studies one worked example.

On the next example, he predicts the elimination choice before it is revealed.

On a third, that decision is blank and Marcus must fill it.

Then he solves a paired problem while the example remains available.

Next, the example is hidden until after his first attempt.

The Fade reduces prompts.

The Independence Test uses a fresh problem after a delay.

Marcus succeeds once.

That is encouraging but not yet stable independence.

A second delayed task is used.

Then a mixed problem requires him to decide whether elimination is even the right route.

Only after repeated unsupported success does the tutor reduce routine intervention.

The Release is not a dramatic graduation from all tuition.

It is release from this specific intensive support condition.

Case 3: The three-student group where one answer hides two learners

This composite case is fictional.

A tutor asks a Science explanation question.

One learner, Wei Jun, answers quickly and correctly.

The other two nod.

The tutor could record "group understands".

The Individual Accountability Gate says that is insufficient.

The tutor asks each learner to write the causal chain privately.

Wei Jun reproduces it.

Sara has the right keywords but reverses cause and effect.

Imran writes only the observation from the question and does not state the mechanism.

One spoken answer hid three states.

The tutor uses the Shared-Error Reteach Gate and realises the errors are not actually shared.

Sara receives a short contrast between two causal directions.

Imran receives an explanation scaffold linking evidence to mechanism.

Wei Jun receives an extension question rather than repeating the basic explanation.

The group reconvenes.

Each learner answers a fresh variant privately before discussion.

The tutor has differentiated without creating three unrelated lessons.

That is orchestration.

Case 4: The parent asking for another weekly session

This composite case is fictional.

A learner's examination result drops.

The parent asks to increase tuition from once to twice a week.

The request is understandable.

The Handbook does not answer immediately.

The Dosage Differential asks whether the present route is insufficiently frequent or simply wrong.

The Aggregate-Score Compensation Trap checks whether the total score hides one component.

It does.

Most components are stable. One comprehension component fell sharply.

The Progress-Review Window Gate examines several recent pieces rather than allowing one examination to dominate.

The tutor sees that the weak component has been inconsistent for several weeks.

The Learning Budget asks where additional time would go.

The Schedule-Substitution Gate asks what a second lesson would displace.

The learner already has a heavy school schedule.

The tutor decides not to add a permanent second weekly session yet.

Instead, the route uses a short temporary practice cycle targeted at the component, with one additional check-in before the next examination.

The Re-Entry Plan defines the exit.

If the component stabilises, cadence returns to normal.

If not, the route is reconsidered.

The parent receives a clear explanation of why "more" was not treated as the first answer.

Case 5: The tutor who attended training but does not use the practice

This composite case is fictional.

A tuition programme introduces a feedback routine.

Tutors attend a ninety-minute workshop.

At the end, almost everyone can describe the routine correctly.

Two weeks later, lesson observations show that several tutors have returned to old correction habits.

A weak programme says:

"They were trained. They are not following instructions."

The Training Need Gate asks what the gap actually is.

Interviews show that tutors understand the routine.

The Training-to-Live-Practice Transfer Check reveals that they struggle to use it while simultaneously managing three learners.

The programme moves from knowledge delivery to rehearsal.

Tutors practise the routine with short case simulations.

The Coaching Focus Gate narrows observation to one move.

The Coaching Receipt checks for live use.

The Implementation Feasibility Gate reveals another problem: the original recording form adds too much documentation after every lesson.

The form is reduced without removing the active ingredient.

The Implementation Acceptability Gate captures tutor feedback about workload.

The Implementation Reach Gate checks whether online tutors and newer staff received the same support.

The Implementation Sustainability Gate later examines whether the routine survives when observation frequency drops.

The programme has now stopped treating a workshop as implementation.

Case 6: The new practice everyone likes

This composite case is fictional.

A tuition centre introduces a colourful digital quiz platform.

Learners enjoy it.

Parents report that children are more willing to practise.

Tutors say lesson energy has improved.

Those are relevant observations.

They do not prove learning improved.

The Implementation Acceptability Gate records the positive response.

The Human-Supported AI Engagement Gate asks whether interaction with the platform produces meaningful learning opportunity or simply more activity.

The Adaptive-Difficulty Comparability Gate notices that the platform changes question difficulty as learners answer.

Percent-correct scores across weeks are therefore not directly comparable without accounting for difficulty.

The Instructional Material Coherence Gate checks whether the generated practice matches the curriculum route.

The Fresh-Item Reserve protects some independent questions outside the platform.

The programme then has a better evidence design.

Enjoyment remains a benefit.

It is no longer asked to carry an effectiveness claim it cannot support.

Case 7: The school and tutor use different methods

This composite case is fictional.

A Primary Mathematics learner is taught one model method in school.

The tutor uses a different representation that the learner finds easier.

At first, performance improves.

Before examinations, confusion appears because the learner cannot map one notation onto the other.

The Classroom–Tuition Translation Gate identifies the risk of parallel systems.

The tutor does not immediately abandon the clearer tuition method.

Instead, the Curriculum-Method Reconciliation Gate maps equivalent steps explicitly.

The learner practises translating between representations.

The Alternative Method Gate checks whether the tuition method is valid and assessable.

The Response-Modality Equivalence Gate is not the owner because the issue is not oral versus written response; it is method representation.

That distinction matters.

The learner eventually understands both representations and knows which one to use under school assessment conditions.

Case 8: The tutor who looks excellent during observation

This composite case is fictional.

A tutor delivers an impressive observed lesson.

Questions are well sequenced.

Feedback is concise.

All three learners participate.

The programme could conclude that coaching is complete.

The Observation Reactivity Check reminds the coach that being observed can change teaching.

The Tutor-Observation Sample Gate asks whether the selected lesson was representative.

The coach samples another ordinary session several weeks later.

The targeted practice still appears.

A third sample during a busy examination period shows partial drift.

The Drift Check identifies which part of the routine disappeared.

The Coaching-Support Taper Gate keeps light support in place rather than either returning to intensive coaching or declaring permanent success.

The evidence now represents ordinary practice more accurately.

Claim the scientific job before writing

Before a new volume is drafted, the editorial process should write one internal sentence:

This article owns the decision of…

If that sentence is vague, the article is not ready.

If the sentence duplicates another owner, the collision must be resolved.

If the sentence describes a broad subject instead of a decision, narrow it.

This is the most important control in the series.

A long article cannot compensate for an unclear job.

Search the live Handbook before assigning an owner

The collision scan should include:

• exact title concepts,
• synonyms,
• adjacent mechanisms,
• older volumes whose language differs,
• and later articles that may already have absorbed the job.

Search should not rely only on volume number.

The live series contains historical number collisions, so job language matters more than the address.

The title should express the decision tension

Strong Handbook titles usually contain two elements:

the named gate, check, boundary, differential, audit, window, protocol or mechanism

and

the tension it resolves

Examples:

"The Support Provenance Check — How a Tutor Interprets Homework When AI, Parents, Peers or Model Answers Helped Produce It Without Treating Supported Work as Independent Evidence."

"The Ratio–Tutor-Skill Fit Gate — How a Tuition Programme Decides Whether Three Learners Can Be Taught as One Group Without Treating 3:1 as Automatically Individualised."

"The Implementation-Acceptability Gate — How a Tuition Programme Uses Tutor, Learner and Family Reactions to a New Practice Without Treating Popularity as Evidence or Dismissing Resistance as Mere Attitude."

The long subtitle is not decoration.

It states the boundary.

The opening should present a real decision, not an abstract definition

Begin with a tutoring event.

The learner gets the answer right after a hint.

Two tutors disagree about the same work.

A parent asks for more sessions.

A coach observes a polished lesson.

A digital platform changes difficulty.

A model essay produces imitation.

A group answer hides individual misunderstanding.

The reader should recognise the job before the article names the mechanism.

Give a quick answer early

Longform should not force a reader to earn the answer by scrolling.

After the setup, state the operating answer in compact form.

Then earn the nuance.

The quick answer should be useful enough for a tutor in the middle of planning.

The rest of the article should explain conditions, exceptions, evidence and implementation.

State the ownership boundary near the top

Name the neighbouring Handbook owners.

Explain the difference.

This reduces cannibalisation and teaches the reader how the network works.

If the distinction cannot be explained clearly, the article may not deserve a separate owner.

Research should inform the job, not dominate the voice

Use current high-quality sources where the job depends on external evidence.

Prefer official guidance, evidence syntheses, practice guides and primary research over unsourced teaching claims.

State dates where guidance changes.

Distinguish classroom evidence from private-tutoring inference.

Do not turn an average effect into a guarantee for one learner.

Do not force a research finding into a context it did not study without saying that the transfer is professional judgement.

The public article should read as a coherent teaching manual, not a literature dump.

Composite cases should test the boundary

A case is useful when it forces the mechanism to make a decision.

Do not add fictional learners as decoration.

Use them where the tutor could reasonably choose two routes and the article shows why one is better supported.

Mark composite cases as constructed.

Protect privacy.

Failure modes belong before the conclusion

Every article should name predictable ways the mechanism can be misused.

This is where many Handbook pages become practically valuable.

A tutor can understand the ideal and still fail because of:

over-support,

under-support,

measurement contamination,

fixed routines,

false precision,

premature causal claims,

group leakage,

poor transfer,

workload,

or role expansion.

Failure modes convert principles into safeguards.

Evidence boundaries should be explicit

End research-heavy articles with a section that says what the evidence does and does not establish.

This is especially important for:

effect sizes,

implementation research,

AI claims,

small-group comparisons,

assessment interpretations,

and psychological mechanisms.

The Handbook should prefer a narrower true claim to a broader impressive one.

The conclusion should return to the decision

Do not end with a generic motivational paragraph.

Return to the gate.

What should the tutor notice?

What should they check?

What should they do?

What would count as a receipt?

When should they move to the adjacent owner?

That is how the article becomes operational.

"The learner is wrong."

Start with diagnosis.

Use the Diagnostic Tutor, Fresh Look, Differential, Diagnostic Probe, Task Purity Check, Tutor-Side Check, Omission Audit, Missing-Response Differential or Response-Time Interpretation Gate depending on what is unknown.

Do not start by adding more practice unless the cause is already clear.

"The learner is right."

Ask whether the success needs auditing.

Use Correct-Answer Audit, Support Provenance, Evidence Freshness, Changed Question, Mixed Set or Independence Test depending on what the answer is supposed to prove.

Do not turn every correct answer into an interrogation.

Audit when the decision cost or conflicting evidence justifies it.

"The learner understands when I help."

Use Dose, Worked-Example-to-Problem Alternation, Fade, Question Return, Follow-Up Question and Independence Test.

Ask what part of the performance the support is carrying.

Transfer one consequential piece at a time.

"The learner can do it but not in the exam."

Use Performance Coach, Timed Set, Full Paper, Second Full Paper, Aggregate-Score Compensation Trap, Progress-Review Window and relevant component owners.

Separate knowledge from performance coordination.

"The group looks fine."

Use Individual Accountability, Peer Answer Leakage, Peer Explanation, Shared-Error Reteach, Learner-Talk Interpretation and Three-Learner Orchestration.

Create at least one individual receipt before trusting group success.

"The learner wants a different method."

Use Alternative Method Gate and Curriculum-Method Reconciliation.

Judge validity, generality and assessment fit before enforcing conformity.

"The parent wants more tuition."

Use Dosage Differential, Learning Budget, Schedule-Substitution, Progress-Review Window and Re-Entry Plan.

Ask whether the issue is amount, route or evidence.

"The tutor needs improvement."

Use Training Need Gate, Coaching Focus, Rehearsal-to-Live, Observation Window, Coaching Receipt and Professional-Learning Cycle.

Do not assume every performance problem is solved by more training.

"The programme wants to roll out a new practice."

Use Evidence Source Gate, Implementation Readiness, Feasibility, Active-Ingredient Adaptation, Reach, Acceptability, Consistency and Sustainability.

Do not move from "promising idea" to "everyone must use it" in one step.

"The data says we improved."

Use Measurement Range, Criterion Drift, Missing Baseline, Programme-Average Boundary, Attrition Evidence Gap, Causal-Attribution Gap and correlated-evidence controls.

Ask what changed, under what conditions, and what the evidence can legitimately support.

"AI can do this faster."

Use AI Material Verification, Support Provenance, Human-Supported AI Engagement, Instructional Material Coherence and Learner-Data Retention.

Speed does not transfer accountability.

The Tutor Handbook is an applied professional decision system.

It draws on evidence-informed teaching principles, tutoring research, formative assessment logic, implementation science, cognitive science, professional learning, measurement reasoning and the accumulated realities of tutoring practice.

External research matters.

So do its boundaries.

The Education Endowment Foundation currently reports positive average effects in its one-to-one tuition and small-group tuition summaries, while emphasising targeting, links to normal teaching, tutor quality, training and progress monitoring. Its small-group summary also notes meaningful variability and warns against assuming precise group size alone determines effectiveness.

Those findings support the Handbook's focus on targeted teaching, diagnosis and tutor quality.

They do not prove that every private tuition programme, every three-student group or every Handbook method will produce the average effect.

AERO's explicit-instruction guidance supports clear modelling, sequenced teaching, guided practice, worked examples and the gradual removal of scaffolding as proficiency grows. Its current guides and implementation resources also emphasise professional learning, context and monitoring implementation outcomes.

Those sources support several mechanisms in the Handbook.

They do not remove the need to inspect the individual learner, local curriculum, tutor capability and actual implementation.

Research averages describe groups under study conditions.

Tutoring decisions concern a particular learner or programme under current conditions.

The Handbook bridges those levels through professional judgement that is explicit, revisable and evidence-seeking.

That is why so many articles contain gates.

The gate is where evidence meets action.

Good tutor judgement is not intuition without evidence.

It is also not a checklist without interpretation.

It has several properties.

It is specific

The tutor can state the decision.

It is evidence-linked

The tutor can state what observation supports the decision.

It is differential

The tutor considered plausible alternatives.

It is proportionate

The action matches the evidence and the cost of being wrong.

It is reversible where uncertainty remains

The tutor prefers a trial when a permanent change is not yet justified.

It preserves learning opportunity

The intervention does not solve so much that the target capability disappears.

It preserves independence as a destination

Support has a direction of travel.

It is communicable

Another tutor, learner or parent can understand why the route changed.

It is updateable

New evidence can change the conclusion without anyone losing face.

It respects scope

The tutor knows when the decision belongs elsewhere.

This is the professional core the two-hundred-volume series has been building.

The Handbook can be compressed into twelve standing rules.

1. Observe before labelling.

Describe what happened before deciding what it means.

2. Define the target capability.

Evidence is uninterpretable until the target is clear.

3. Separate competing explanations.

Do not let the first plausible story become the diagnosis.

4. Collect evidence that changes the decision.

More data is not automatically better evidence.

5. Choose the smallest useful intervention.

Large changes should earn their cost.

6. Record the support condition.

A performance with help is not the same evidence as the same performance without it.

7. Transfer responsibility.

Teaching should progressively make the learner more capable of carrying the route.

8. Verify under a changed condition.

Freshness, delay, changed surface, mixed methods or reduced support can reveal whether learning travels.

9. Precommit review rules where practical.

Decide what would cause a route change before the next result arrives.

10. Preserve boundaries.

Between teaching and doing the work; support and substitution; relationship and role expansion; data and identity; adaptation and drift.

11. Make programme quality observable.

Training, coaching and implementation should produce receipts in live practice.

12. Say only what the evidence supports.

A narrow accurate claim is stronger than a broad impressive one.

The Tutor Handbook began with ordinary questions.

What kind of tutor does this learner need?

What should happen in the first lesson?

How should practice work?

How do we know improvement is real?

When should support fade?

Those questions opened into a larger system.

Once tutoring takes learner evidence seriously, it must care about measurement.

Once it cares about measurement, it must care about support conditions, comparability, task purpose and causal claims.

Once it cares about teaching quality, it must care about tutor training, observation and coaching.

Once a programme wants quality across tutors, it must care about implementation, consistency, feasibility, reach, acceptability and sustainability.

Once technology enters the room, it must care about AI verification, provenance, data ownership and material coherence.

Once several learners share a tutorial, it must care about orchestration, accountability, grouping and case mix.

The series therefore grew naturally from tutoring technique into a complete professional operating system.

That growth should continue only when a new decision genuinely deserves an owner.

The root manual remains stable in purpose even as new volumes appear.

Its job is to keep the system legible.

For the tutor at the table, the operating question remains simple:

What does this learner need next, what evidence supports that judgement, what is the smallest useful action, and what fresh evidence will tell us whether it worked?

For the coach, the question becomes:

What teaching move needs to change, what evidence would show live transfer, and how can support taper without the practice collapsing?

For the programme leader:

What conditions make good tutoring repeatable across people and time without replacing professional judgement with scripts?

For the learner:

What can I now understand, choose, perform and regulate that I previously needed someone else to carry?

That final question is the reason the Handbook exists.

A tutor does not succeed by becoming permanently indispensable.

A tutor succeeds when explanation becomes understanding, guided practice becomes independent selection, feedback becomes self-correction, supported performance becomes durable capability, and the learner can carry more of the route into the future.

The Handbook is the operating manual for making those transitions deliberately.

The article should survive a five-level review

A Tutor Handbook volume should not be judged only by whether the prose is clear.

It should survive five different tests.

Job test

Can an editor state the article's unique decision in one sentence?

If not, the article is not ready.

Evidence test

Do the claims match the quality and population of the evidence?

If research concerns classroom teaching, the article should not quietly claim direct proof for private one-to-three tuition unless that transfer is justified.

If the article uses a practice guide, it should identify the guide as guidance rather than a direct trial of the exact tutoring system.

If the article uses programme data, it should not turn local improvement into a universal claim.

Collision test

Does another Handbook owner already answer the same decision?

Similarity of topic is acceptable.

Duplication of owner job is not.

Execution test

Can a tutor actually use the article in a lesson, review, coaching conversation or programme decision?

A strong article should contain recognisable triggers, diagnostic questions, actions and receipts.

Boundary test

Does the article protect against predictable overreach?

The article should say what the mechanism cannot prove, when it should stop, what neighbouring owner takes over and what failure modes would contaminate the result.

Passing all five tests makes the page durable.

Quality should be visible in the architecture

A long article can hide weak reasoning behind volume.

The Handbook should therefore make its reasoning visible.

The reader should be able to find:

the job,

the quick answer,

the evidence required,

the key decision rule,

the main exceptions,

the failure modes,

the evidence boundary,

and the adjacent owners.

This architecture is not merely for search.

It is for professional usability.

A tutor may first read an article in full.

Months later, the same tutor may return during planning and need only the decision rule.

The article should work in both modes.

Longform should earn its length

The Handbook uses longform where the job requires it.

A 20,000-word article is justified when the decision involves several mechanisms, multiple stakeholder perspectives, evidence boundaries, worked cases, implementation conditions and a durable reference function.

It is not justified by a numerical target alone.

Substantive length should come from:

definition,

causal mechanism,

diagnosis,

worked examples,

counterexamples,

failure modes,

transfer,

measurement,

implementation,

professional boundaries,

research context,

FAQs where useful,

and navigation to adjacent owners.

Do not inflate length with repeated summaries, decorative quotations, generic motivation or keyword variants.

The root manual is long because it controls the system.

A narrow gate may need far less.

Function determines scale.

Search language should follow reader intent

Tutor Handbook titles are often technical because the named mechanism helps preserve ownership.

The article should still contain the language a real tutor, parent or programme lead might use.

A reader may search:

"student understands in tuition but cannot do alone",

"how much help should a tutor give",

"when to change tutor",

"how to know tuition is working",

"student gets correct answer but doesn't understand",

"small group tuition three students",

"AI homework how can tutor tell",

"how to train tutors",

"how to observe a tutor",

or

"more tuition or wrong tuition".

The article can answer natural-language intent while retaining a precise canonical owner.

Search optimisation should therefore improve discoverability without changing the job.

The job remains primary.

Internal links should describe decisions

"Read more" is weaker than a decision link.

Prefer:

"If the problem is that the learner can perform only while prompts remain, move to the Fade."

"If the result is correct but may have been produced by external help, use the Support Provenance Check."

"If the practice is strong in training but absent in live lessons, use the Training-to-Live-Practice Transfer Check."

This makes the internal link part of the tutoring logic.

The website becomes a navigable decision system rather than an archive.

Gate

A gate determines whether conditions are sufficient for a transition.

Examples include moving to harder work, adopting a practice, changing a tutor, reducing support or launching implementation.

A gate should state the entry evidence and the consequence of passing or not passing.

Check

A check inspects whether an assumption, result or condition is reliable enough for the next decision.

Checks are often lighter than gates.

They may confirm implementation fidelity, evidence freshness, opportunity to demonstrate a skill, or whether the tutor's own teaching created the problem.

Boundary

A boundary separates two legitimate activities that become harmful when blurred.

Examples include support versus substitution, preparation versus predetermination, professional warmth versus role expansion, extension versus acceleration, and access support versus performance of the target skill.

Differential

A differential separates plausible causes that would require different actions.

It is borrowed as a reasoning pattern, not as a medical claim.

The tutor does not merely list possibilities.

The tutor seeks evidence that distinguishes them.

Audit

An audit reopens an apparently settled result.

A correct answer can be audited.

An omission can be audited.

The goal is not suspicion for its own sake.

It is to inspect cases where the visible outcome may hide a consequential difference.

Window

A window defines the period or sample of evidence that should inform a decision.

Too narrow a window overreacts to noise.

Too wide a window allows old evidence to dominate current performance.

Protocol

A protocol structures a recurring interaction in which unstructured judgement is likely to drift.

Examples include disagreement between tutor and coach.

The protocol should support evidence use without replacing professional responsibility.

Trap

A trap names a predictable inference error.

Regression to the mean, correlated evidence, peer baselines, aggregate-score compensation and subscore resolution are examples.

Naming the trap gives the tutor a cognitive warning before a decision is made.

Receipt

A receipt is observable evidence that the intended change occurred.

For learner feedback, the receipt is changed performance.

For coaching, the receipt is changed live teaching.

For implementation, the receipt may be reliable use under ordinary conditions.

A receipt is stronger than agreement or intention.

Route

The current working plan linking target, evidence, hypothesis, intervention, support, receipt and review rule.

A route should be coherent enough to follow and flexible enough to change when evidence changes.

Tutors

Do not attempt to read the series from Vol. 0001 to the frontier before using it.

Start with the current decision.

Use the root router.

Read the owner article.

Follow adjacent owners only if the evidence moves the case there.

Over time, the network becomes familiar.

New tutors

Begin with the early role and lesson architecture.

Understand explanation, diagnosis, practice, performance, the first lesson, review, learning contract and handover.

Then learn the core evidence and support sequence:

Fresh Look,

Differential,

Branch Point,

Dose,

Fade,

Independence Test,

Learning Claim.

These provide the grammar needed to use later gates well.

Experienced tutors

Use the later volumes to inspect invisible habits.

Experienced tutors often possess strong routines.

The risk is that successful routines become assumptions.

Use Tutor-Side Check, Disconfirmation Check, Criterion Drift, Observation Reactivity, Alternative Method Gate and correlated-evidence controls to keep judgement updateable.

Tutor coaches

Use the Handbook to turn vague coaching into observable jobs.

Instead of "improve questioning", identify whether the tutor needs better diagnostic probes, follow-up questions, question return, wait conditions, or individual accountability in groups.

Then define a live receipt.

Programme leaders

Use the Handbook at two levels.

Learner-level owners protect instructional quality.

Programme-level owners protect staffing, coaching, implementation and consistency.

Do not confuse the two.

A programme average cannot diagnose an individual learner.

An individual success cannot establish programme reliability.

Parents

The Handbook can help parents ask better questions.

Not "How many worksheets did my child finish?"

Ask:

"What are you currently trying to improve?"

"What evidence tells you that is the bottleneck?"

"What can my child now do without help?"

"What will make you change the plan?"

"How will we know when this support is no longer needed?"

These questions encourage transparency without asking parents to become the tutor.

Learners

Learners can use the language of routes and receipts too.

"What is my current job?"

"What help am I allowed to use?"

"What will I try alone?"

"What result would show I am ready to move on?"

"What should I do if the old problem returns?"

A learner who can answer these questions is gaining control over learning, not merely completing tuition.

The Handbook now reaches Vol. 0210 and may continue beyond it.

The frontier is not a finish line.

Future volumes should be added only when they extend the operating system rather than crowd it.

A future owner may emerge because:

a new technology changes the provenance of work,

a curriculum reform creates a new translation problem,

a recurring coaching failure becomes visible,

a measurement condition needs its own control,

a programme implementation issue recurs,

or a boundary currently buried inside another article proves consequential enough to stand alone.

Before publication, ask:

What decision becomes easier or safer because this page exists?

If the answer is merely "the reader knows more about the topic", the job is not sharp enough.

If the answer is "the tutor can now distinguish two situations that require different actions", the owner may be justified.

Preserve the root as the control page

As new volumes publish, this root should be updated in three ways.

First, extend the live frontier and registry.

Second, place the new volume in the decision architecture.

Third, check whether the new owner changes any root routing language.

Do not rewrite the root simply to chase every new article.

The root should remain a stable manual.

Its architecture changes only when the system itself changes.

Keep historical integrity

Public URLs are part of the system's history.

Do not casually delete old articles because a later owner is better.

Where a true collision exists, choose deliberately among:

retaining both with clarified boundaries,

merging and redirecting,

turning one into a companion casebook,

or marking one as superseded.

The decision should preserve reader routes and canonical clarity.

The Handbook should become easier to use as it grows

Growth is successful only if navigation improves faster than complexity.

Every new owner should therefore strengthen at least one route.

A new article that creates three new ambiguities for one solved decision has not improved the system enough.

The root, registry, adjacent-owner links and job contract together keep growth controlled.

Tutoring often appears small from the outside.

One tutor.

One learner or a few learners.

A table.

A worksheet.

A lesson.

But the professional decisions inside that small setting are not small.

The tutor is constantly deciding what evidence means, what to teach, what not to teach yet, how much help to give, whether a success is real, whether a failure matters, whether the learner should struggle a little longer, whether the route should change, whether a parent concern changes the plan, whether an AI-generated resource is trustworthy, whether a group answer hides individual weakness, whether a method should be reconciled with school, whether coaching has transferred, whether an implementation problem belongs to the tutor or the system, and whether continued tuition is still creating enough value to justify its cost.

A handbook that takes those decisions seriously must do more than provide tips.

It must make professional judgement inspectable.

That is the job of The Tutor Handbook.

Each volume owns one decision.

The root connects the decisions.

The tutor uses evidence to move through them.

The learner should emerge with more capability and more control than they had before.

That is the standard.