Three-Student English Tutorials | Visibility, Feedback and Independent Performance

The teaching model: three is a design constraint, not a promise

A three-student tutorial should not be sold as if the number itself produces better learning. Three students can still sit through a lecture. Three students can still copy model answers. Three students can still become dependent on a tutor who intervenes too quickly. A small group only creates conditions that a teacher can use well or badly.

The important question is therefore not “Is three better than eight?” in the abstract. It is: What becomes possible when a teacher is responsible for only three visible decision processes at a time, and is the lesson designed to use that visibility?

SETC’s answer is that the format is most valuable when English work contains hidden decisions that need to be observed, compared, repaired and then released back to the learner. Reading, writing, oral communication and examination performance all contain such decisions. A final answer alone often hides them.

What the number three changes

Three students create a particular observation problem. The tutor has to track three learners, but the learners also become observable to one another. That creates a triangle rather than a tutor–student line.

In that triangle, several useful things can happen:

• the tutor can still see individual work closely;
• students can attempt the same task independently before discussion;
• different reasoning paths become visible side by side;
• one learner’s error can become a contrast case for the others without becoming a public humiliation;
• the tutor can ask one student to explain while observing whether the other two can evaluate the explanation;
• feedback can stay student-specific even when the explanation is shared;
• students can hear questions they did not think to ask;
• the tutor can fade support and watch whether independence survives while still having two comparison cases.

That last point matters. In one-to-one teaching, a student’s successful answer can be difficult to interpret because the tutor may have supplied subtle cues. In a three-student tutorial, the teacher can compare what happened under similar conditions across three attempts. The comparison does not prove causation, but it gives more diagnostic information.

What the number three does not change

Three students do not remove the need for curriculum knowledge, language expertise, task design or feedback skill. They do not guarantee compatible personalities. They do not make every student equally talkative. They do not solve weak motivation automatically. They do not make travel, scheduling or cost irrelevant.

Most importantly, three students do not guarantee individualisation. A tutor who gives the same explanation, same worksheet and same correction to all three is running a tiny class, not necessarily a diagnostic tutorial.

The format is strongest when the work is visible

SETC uses visibility as the first design principle. Visibility means that the tutor can observe enough of the learner’s process to infer where the performance began to break.

For reading, visibility can include:

• which sentence the student identifies as evidence;
• which pronoun they think refers to which noun;
• where an inference begins;
• how much text they treat as relevant;
• how they paraphrase the question;
• when they decide to stop searching.

For writing, visibility can include:

• how the prompt is interpreted;
• which ideas are selected or rejected;
• whether the student defines a paragraph job before drafting;
• how examples are connected to claims;
• where sentence-level editing begins to consume too much time;
• what the student checks before submission.

For oral communication, visibility can include:

• how quickly the learner forms an answer;
• whether they answer the exact question or a nearby one;
• how they develop a reason;
• what happens after interruption or follow-up;
• whether pronunciation difficulty affects intelligibility;
• how they repair when a listener looks confused.

For examinations, visibility can include:

• time spent per item;
• which questions trigger rereading;
• when checking becomes unproductive;
• whether one hard item disrupts later work;
• how the student recovers after uncertainty.

These are process observations. They are more useful than broad labels such as “weak comprehension” or “careless writing”.

The first repairable cause

A visible error can be several steps downstream from the actual cause. Suppose a student loses a comprehension mark because the final sentence is too broad. The tutor could correct the wording. But the first repairable cause may have occurred earlier: the student treated the entire paragraph as evidence when the question required one contrast.

SETC therefore traces backwards:

final answer → language form → selected evidence → inference → question interpretation → passage model.

The first point where the learner’s model becomes unstable is often the best teaching target.

This does not mean every error needs a deep diagnostic excavation. Some errors are simple. The principle is to avoid repairing the last visible symptom automatically when an earlier repeated decision is producing several symptoms.

Three students make divergence visible

Imagine one inference question.

Iona chooses the correct sentence but interprets one pronoun incorrectly.

Maren understands the pronoun but uses evidence outside the requested scope.

Leonie understands both but changes a correct answer after excessive checking.

The same lost mark would conceal three different causes. A generic explanation of the answer would be inefficient because each student needs a different repair.

In a three-student tutorial, the tutor can preserve the shared task while individualising the diagnosis.

One-to-one, three-student and larger-group teaching solve different problems

SETC does not need to claim that one class size is universally superior. Different formats create different instructional trade-offs.

One-to-one teaching

One-to-one teaching maximises tutor attention and can be valuable when a learner needs privacy, highly specialised pacing or intensive repair. The tutor can stop immediately and redesign the lesson around one student.

But one-to-one teaching also has limitations:

• the learner sees only their own approach unless the tutor manufactures alternatives;
• the tutor–student interaction can become overly cue-dependent;
• the learner may begin performing for the tutor rather than for the task;
• discussion lacks genuine peer contrast;
• the teacher can accidentally rescue too much because every hesitation is highly visible.

Those are not arguments against one-to-one teaching. They are design risks.

Three-student tutorials

Three students preserve substantial individual visibility while introducing peer contrast. The tutor can ask one learner to explain, another to challenge the evidence and the third to evaluate which answer best fits the task.

The format is especially useful when the objective is not merely acquiring information but learning to make decisions independently.

Larger small groups

Larger groups can create richer discussion, more varied examples and lower per-student cost. They may work well for explicit instruction, collaborative projects, discussion and practice where individual diagnosis is not required every minute.

As group size increases, however, the tutor has less bandwidth to observe every decision path continuously. Students can hide behind group success. A correct answer from one learner can mask confusion in several others.

The solution is not to declare larger groups bad. It is to match teaching method to group size.

Why peer contrast matters

Peer contrast is not peer correction without supervision. The purpose is to expose alternative decisions.

Suppose three students plan an argumentative paragraph.

• Maren chooses a strong claim but weak evidence.
• Iona chooses excellent evidence but cannot explain the warrant.
• Leonie chooses a modest claim that fits the evidence and can be completed under time.

The tutor can ask: Which plan is most defensible? What would need to change in each? The students learn to evaluate writing before drafting.

This is more transferable than being shown one model paragraph after the fact.

Contrast reduces the illusion of one correct method

Many English tasks allow several successful approaches. Two strong introductions can differ. Two readers can annotate differently. Two speakers can structure a response differently.

Seeing multiple successful approaches helps learners distinguish principles from templates.

The principle may be “state the position clearly enough that later evidence has a job”. The template “always begin with a rhetorical question” is not the principle.

Contrast also reveals false flexibility

Not every alternative is equally good. Peer comparison should not end with “everyone has their own style”. Students need criteria.

For a comprehension answer, criteria may include evidence, inference and scope. For writing, relevance, structure, support and clarity. For oral, intelligibility, development and responsiveness.

The tutor’s role is to make those criteria explicit enough that students can eventually apply them without the tutor.

Visibility must remain psychologically safe

A small group can make mistakes feel exposed. If every error becomes a public performance, students may hide uncertainty, wait for others or avoid risk.

The tutor therefore needs a safety rule: make the decision visible without making the learner the problem.

Instead of:

“Maren always overthinks.”

use:

“This answer changed from correct to less precise during checking. Let’s examine what evidence triggered the change.”

The unit of analysis is the decision, not the identity.

Stable names, unstable states

SETC’s resident students—Maren, Iona and Leonie—are useful because readers can follow recurring examples. Their names remain stable; their weaknesses should not. Maren can make a strong structural decision in one lesson and a poor one in another. Iona can infer accurately in one text and overreach in another. Leonie can manage time well until a new paper format increases load.

This prevents the examples from teaching fixed-mindset labels.

The tutor’s observation budget

Even with three students, a tutor cannot observe everything at once. Each task should define what is being watched.

A reading lesson might focus on evidence selection.

A writing lesson might focus on paragraph jobs.

An oral lesson might focus on answer development after a follow-up question.

An examination lesson might focus on skip–return decisions.

Trying to diagnose grammar, vocabulary, reasoning, timing, handwriting, motivation and confidence simultaneously creates noise.

Small groups make diagnosis possible; task design makes diagnosis useful.

The visibility rule

Before a tutorial activity begins, the tutor should be able to answer:

1. What decision will each student have to make?
2. What evidence will reveal that decision?
3. What different failure patterns are plausible?
4. What will I do if the task is too easy to reveal anything?
5. How will the learner repeat the decision after feedback?
6. How will I know whether the repair transfers?

If those questions have no answer, the activity may be practice, discussion or instruction—but it is not yet a diagnostic tutorial task.

The three-student operating system: task, observation, intervention, transfer

A three-student tutorial needs an operating rhythm. Without one, the session drifts between teacher explanation, worksheet completion and ad hoc correction. SETC uses a repeatable sequence that can be adapted to age, skill and examination context:

independent attempt → visible decision → contrast → diagnosis → narrow intervention → individual retry → recombination → delayed transfer.

The sequence matters because each stage answers a different question.

Independent attempt

The first attempt should happen before discussion when possible. This preserves evidence about what each learner can do without peer or tutor cues.

If students discuss first, the tutor loses diagnostic information. A correct group answer may hide the fact that one learner had no independent method.

Visible decision

The tutor needs an artefact: underlined evidence, a plan, a spoken explanation, a timing mark, a sentence rewrite, a choice between two interpretations. The artefact should expose the target decision.

Contrast

Students compare answers or methods. The comparison should be criterion-based rather than popularity-based. “Which answer is best?” is weaker than “Which answer stays inside the question’s scope and uses the strongest evidence?”

Diagnosis

The tutor identifies the first weak link. Sometimes all three students share it; sometimes each has a different failure.

Narrow intervention

Teach the smallest thing that can change the next attempt. If the problem is pronoun reference, do not lecture on the entire comprehension paper. If the problem is paragraph purpose, do not rewrite the whole essay.

Individual retry

Each student repeats the decision independently. The tutor should not accept a group chorus as evidence that the repair worked.

Recombination

The repaired skill returns to a more complete task. A student who fixed evidence selection on one question must use it inside a full passage. A student who improved paragraph structure must draft the paragraph under realistic conditions.

Delayed transfer

At a later point, use a different surface task. The learner should not know that the same mechanism is being tested. This is the strongest evidence that the skill has become more portable.

Task design: one task, three visible pathways

The most productive three-student tasks have a common core and individual evidence.

Examples:

• all three read the same passage but submit separate evidence maps;
• all three respond to the same prompt but write separate one-sentence theses before discussion;
• all three hear the same audio but record different notes independently;
• all three answer the same oral question but take turns responding without hearing the previous answer first;
• all three complete the same timed section but record where they skipped, returned and checked.

The shared task makes comparison meaningful. The individual artefact preserves diagnosis.

Variation after repair

Once a weak link is identified, the second example should not be identical. If the learner succeeds only when the noun positions, sentence length and question wording remain the same, the repair may be surface-bound.

Change one dimension at a time at first, then several.

Intervention thresholds

A tutor who interrupts every error destroys independence. A tutor who never interrupts can allow misconceptions to consolidate. The decision to intervene needs criteria.

Intervene when the learner is practising the wrong relationship repeatedly

If Iona repeatedly selects evidence based on keyword matching rather than meaning, another ten questions may reinforce the wrong method. Stop and repair.

Wait when the learner is still productively testing options

Maren pauses before choosing a paragraph order. She compares two structures and rejects one. The hesitation is doing useful cognitive work. Do not rescue it merely because silence feels inefficient.

Intervene when one error will contaminate the rest of the task

If a student misunderstands the central premise of a passage, later answers may all be invalid. A short correction protects the remaining practice.

Wait when the learner can detect the error through checking

If Leonie has written a sentence with a reference ambiguity but the lesson target is self-editing, give her the chance to find it.

Intervene when emotional state blocks learning

Productive struggle has a boundary. Confusion that becomes panic, shutdown or humiliation no longer produces useful evidence. Reset the task, reduce load and rebuild.

Three kinds of tutor talk

Small groups become teacher-dominated easily. SETC distinguishes three kinds of tutor talk.

Explanation

The tutor supplies a concept or model the learner does not yet have. Explanation should be concise enough that students return quickly to doing.

Questioning

The tutor exposes the learner’s model:

• What made you choose that sentence?
• Where does the question limit the answer?
• What is this paragraph trying to make the reader understand?
• What would change your conclusion?
• Why did you return to that question?

Questions are diagnostic when they reveal thinking, not when they lead the learner toward one predetermined answer word by word.

Feedback

The tutor identifies the effect of a decision and sets the next action. Good feedback reduces future uncertainty.

Role rotation

Three learners allow temporary roles that focus attention. Roles should rotate so no student becomes “the reader” or “the weak writer”.

Evidence keeper

Checks whether claims are supported by the text or source.

Scope keeper

Checks whether the answer stays inside the question.

Clarity keeper

Checks whether the final language makes the reasoning recoverable.

On the next task, rotate.

Writer, challenger, verifier

One learner proposes a thesis. Another gives the strongest reasonable challenge. The third checks whether the evidence can sustain either position.

Then rotate.

Speaker, listener, observer

One answers an oral question. One listens for meaning and asks a follow-up. One observes structure and intelligibility. Then rotate.

These roles make peer participation functional rather than decorative.

Peer explanation

Explaining can strengthen the explainer’s own model because it forces relationships into language. But peer explanation should not be treated as automatically correct.

The tutor listens for:

• missing steps;
• false rules;
• overgeneralisation;
• ambiguous terms;
• examples that do not match the rule.

A peer explanation is evidence of understanding and an opportunity for diagnosis.

Peer correction

Students should not become substitute teachers correcting one another constantly. Peer correction works best when the criterion is narrow and visible.

Examples:

• Does this answer include evidence outside the scope?
• Can you identify the referent of every pronoun?
• Which sentence is the paragraph’s main claim?
• Did the speaker answer the follow-up question?

Narrow criteria reduce unproductive disagreement.

Wait time

Small-group lessons can move too quickly because the tutor can see all three students. Wait time is still necessary.

A learner may need several seconds to retrieve a word, evaluate evidence or reframe a sentence. If the tutor supplies the answer immediately, the learner practises recognition rather than retrieval.

Participation equity

Equal speaking time is not always the goal, but equal access to cognitive work is.

A confident student can dominate by answering first. Use structures that preserve independent thought:

• everyone writes before anyone speaks;
• randomise response order;
• ask the second speaker to respond to the first rather than repeat;
• use silent planning time;
• require each student to produce one independent artefact.

The fast student

A fast learner can become bored if every task ends when the slowest student finishes. Do not reward speed with more of the same worksheet.

Extension can increase depth:

• defend an alternative interpretation;
• create a counterexample;
• rewrite for another audience;
• explain why a tempting answer is wrong;
• generate a harder question;
• identify what would make the current method fail.

This keeps the learner inside the same mechanism while increasing complexity.

The slower student

Slower processing does not necessarily mean weaker understanding. The tutor should determine whether time is being spent on useful reasoning, retrieval difficulty, decoding, uncertainty or repeated checking.

Support can change the task load without giving away the decision.

The quiet student

Quiet participation can reflect temperament, uncertainty, language formulation load or group dynamics. Do not infer lack of thinking from low volume.

Use written preparation, turn order and specific invitations. Evaluate the quality of decisions, not extroversion.

The dominant student

A dominant student may answer before peers have thought, correct others prematurely or convert discussion into performance. The tutor needs boundaries:

• silent first attempt;
• assigned response order;
• roles that require listening;
• questions that ask the student to evaluate another method rather than produce another answer.

The dependent student

Some learners look toward the tutor after every sentence. In a tiny group this behaviour can intensify because the tutor is always available.

Use a support ladder:

1. What does the task ask?
2. What have you already tried?
3. Which evidence is available?
4. What are your two best options?
5. Choose one and test it.

The tutor gives help only after the learner has externalised the decision.

Error ownership

Students should revise their own work. If the tutor rewrites the sentence, the tutor owns the repair.

A better sequence:

1. identify the problem type;
2. explain its effect;
3. ask the student to propose a revision;
4. compare alternatives;
5. test the revision in context.

Feedback bandwidth

Three students can receive more feedback than they can use. The tutor should prioritise.

Choose one or two high-leverage targets per task. A writing draft may contain ten grammar errors, but if the paragraph has no controlling idea, sentence polishing is not the highest-return intervention.

Feedback timing

Immediate feedback is useful when an error would reinforce a misconception. Delayed feedback can be useful when the learner needs to complete the process independently before review.

Timing is part of teaching design.

Scaffolds

Scaffolds can include:

• question checklists;
• evidence tables;
• paragraph job prompts;
• sentence frames;
• timing checkpoints;
• model annotations;
• oral response structures.

A scaffold should make the target decision possible, not perform it automatically.

Fading scaffolds

Every scaffold needs a removal plan. A student who always writes with a six-box planner has learned to write with a six-box planner.

Fade by:

1. removing one prompt;
2. changing the order;
3. using an unfamiliar task;
4. reducing tutor checking;
5. delaying feedback;
6. asking the learner to generate their own checklist.

The tutor should become less visible over time

At the beginning of a repair, the tutor may model heavily. Later, the tutor should shift from explaining to questioning, then from questioning to observing.

The final state is not a silent tutor in every lesson. It is a learner who can perform the decisions without needing the tutor to initiate them.

Lesson architecture example: 90 minutes

A 90-minute tutorial does not need a fixed template, but one possible architecture is:

• 10 minutes: retrieval or delayed transfer from previous work;
• 20 minutes: independent diagnostic task;
• 15 minutes: comparison and diagnosis;
• 15 minutes: targeted teaching and narrow repair;
• 20 minutes: recombined task under more realistic conditions;
• 10 minutes: reflection, next-task rule and independent follow-up.

The proportions change with age and task. The important feature is that explanation does not occupy the entire lesson.

Lesson architecture example: younger primary learner

Younger learners need shorter cycles. A lesson might alternate oral language, shared reading, movement, writing and game-like retrieval while preserving the same diagnostic logic.

The format should not force young children into prolonged seated comparison merely because three students are present.

Lesson architecture example: JC or advanced learner

Advanced learners can sustain longer independent work and discussion. A GP tutorial might include an unseen source set, individual claim mapping, cross-student synthesis, one focused mini-lesson on evidence calibration, and a timed paragraph completed independently.

Homework in a three-student system

Homework should create evidence for the next lesson rather than merely increase volume.

Useful homework:

• one delayed transfer task;
• one reading annotation with a named criterion;
• one revised paragraph plus explanation of the change;
• one oral recording with self-evaluation;
• one vocabulary retrieval set used in original sentences.

Unreviewed bulk worksheets weaken the diagnostic model because they generate more data than anyone uses.

How the three-student model changes across English skills

The teaching model stays constant, but the visible evidence changes by skill. Reading exposes evidence selection and inference. Writing exposes planning and organisation. Grammar exposes relationships inside sentences. Vocabulary exposes retrieval and collocation. Listening exposes real-time meaning recovery. Speaking exposes formulation and repair. Examination work exposes timing and decision control.

A strong three-student tutorial does not apply one discussion routine to all of them.

Reading: compare evidence before comparing answers

Reading discussions often become answer-sharing. One student says the correct answer and the others agree. That produces little diagnostic value.

SETC instead asks students to preserve the evidence trail.

Step 1: interpret the question independently

Each student writes what the question is asking in plain language. This reveals command-word or scope errors before the passage is searched.

Step 2: select evidence independently

Students mark the exact sentence, phrase or region that supports the answer. They should not discuss yet.

Step 3: compare evidence

If the students selected different regions, the tutor asks why. This is often more informative than comparing final wording.

Step 4: state the inference

Each learner explains what must be inferred from the selected evidence and what remains explicitly stated.

Step 5: control answer scope

Students compare which information the question requires and which information, though true, does not belong.

This approach aligns with the Reading & Comprehension system.

Worked reading example

A passage describes a student who checks the clock twice, rereads the same sentence and then closes the book without writing.

The question asks: What suggests that the student is struggling to begin the task?

Iona selects “rereads the same sentence”. Maren selects “closes the book”. Leonie selects both and adds “checks the clock twice”.

The tutor does not immediately tell them the mark-scheme answer. Instead:

• Which evidence most directly indicates difficulty beginning?
• Which evidence could also occur for other reasons?
• How much evidence does the question require?
• What inference is safe?

The students learn evidence weighting, not answer imitation.

Comprehension vocabulary in context

When a passage word is unfamiliar, the three students can produce independent context hypotheses.

One uses morphology. One uses surrounding contrast. One uses sentence grammar. The tutor can show how several clues converge.

The goal is not to reward the lucky guess. It is to identify the most reliable evidence path.

Summary

Summary work benefits from comparison because selection errors become visible quickly.

Each student identifies the five most important propositions. Then compare:

• What did all three keep?
• What did only one student keep?
• Which detail is an example rather than a central point?
• Which two propositions can be combined without losing meaning?

Compression becomes an explicit decision process.

Writing: compare plans before prose

Writing is expensive to repair after 800 words. Three-student tutorials should expose decisions before drafting.

Prompt interpretation

Each student identifies:

• topic;
• task;
• audience where relevant;
• scope;
• required form;
• important constraints.

Differences are resolved before writing begins.

Idea selection

Students generate more ideas than they will use. The tutor asks them to select according to relevance and explanatory potential, not novelty alone.

Structure

Each student writes paragraph job labels. Compare structures rather than sentences.

Maren may propose:

1. define the problem;
2. explain cause;
3. show consequence;
4. evaluate solution.

Iona may organise by competing positions.

Leonie may choose a more compact structure that can be completed reliably under time.

All can be valid if they answer the task.

Drafting

Students draft independently. The tutor resists line-by-line correction during the first attempt unless a repeated error prevents meaning.

Feedback

Choose one high-level and one language-level target where possible. Example:

High-level: the paragraph does not explain why the evidence matters.

Language-level: pronoun reference becomes unclear when two groups are discussed.

The student revises both.

Narrative writing

Three-student comparison is especially useful for plot causation. Give the same story prompt. Ask each student to produce:

• character goal;
• obstacle;
• decision;
• consequence;
• change by the end.

Students see that an interesting story depends less on adjective density than on controlled cause and consequence.

Situational writing

Assign each student one control layer:

• role/audience;
• source content;
• purpose/action.

After individual planning, they audit one another’s plan, then each writes independently.

Rotate roles on the next task.

Argumentative and expository writing

Peer contrast can reveal reasoning gaps before prose hides them.

One student states the claim.

One gives the strongest counterexample.

One identifies what evidence would distinguish the two.

Then all three write their own paragraph.

This links to Reasoning, Argument & Knowledge rather than treating argument as a paragraph formula.

Grammar: make the relationship visible

Grammar tutorials should avoid turning three students into three people completing the same correction worksheet silently.

Use contrastive examples where students explain meaning differences.

Reference

Give a sentence with two possible antecedents. Each student rewrites it to remove ambiguity in a different way.

Tense and aspect

Give the same event and ask students to describe it as completed, ongoing, habitual and relevant to the present. Compare how grammar changes the timeline.

Modality

Students rank may, might, should, must, will by the commitment they create in a specific context. Then test the choices in real sentences.

Sentence combination

Give several simple propositions. Each student combines them using different clause relationships. Compare what becomes foregrounded, subordinated or ambiguous.

The underlying owner is Grammar & Sentence Control.

Editing

Editing becomes more diagnostic when students have different jobs.

One pass checks meaning relationships.

One pass checks grammar.

One pass checks economy and readability.

Then students edit their own work using all three lenses.

Vocabulary: retrieval before recognition

Three students can create excellent vocabulary contrast because word knowledge is multidimensional.

For one target word, ask each student for:

• meaning in their own words;
• one collocation;
• one grammatical pattern;
• one near-synonym and difference;
• one original sentence;
• one context where the word would be inappropriate.

If one student can define the word but cannot collocate it, the weakness is productive use rather than recognition.

Vocabulary peer challenge

Maren uses the word in a sentence.

Iona asks whether the collocation is natural.

Leonie checks whether a simpler word would be more precise.

Then verify through Corpus, Dictionary & Reference Literacy when needed.

Spelling

Spelling differences should be diagnosed by type:

• sound–letter mapping;
• morphology;
• word-family pattern;
• visual memory;
• national spelling convention.

Three students can compare strategies, but each must retrieve the spelling independently later.

Listening: protect the real-time signal

Listening cannot be taught as if learners can pause the world whenever meaning is lost.

A three-student listening cycle can use:

1. independent first listen;
2. individual notes;
3. comparison of what each listener treated as important;
4. second listen with a specific target;
5. evidence-based correction;
6. new audio for transfer.

Different listening failures

Iona hears the words but misses speaker attitude.

Maren understands the main idea but records too much detail and loses the next section.

Leonie misses one number and keeps thinking about it, damaging the following answers.

The same score can hide three different repairs.

Listening with visuals

One student tracks the audio, one initially studies the visual organisation, one focuses on question cues. Compare how attention allocation affects performance. Then all students repeat individually.

This links to Listening & Viewing and Evidence, Data & Multimodal English.

Speaking: preserve independent formulation

Speaking practice can become imitation if students hear two answers before giving their own. Use preparation controls.

Simultaneous planning

Give the question and allow all three thirty seconds to write only keywords. Then choose response order randomly.

Follow-up questions

The second stage should change. Ask an unseen follow-up that requires adaptation rather than memorised continuation.

Peer listener

One learner listens for the main answer. Another listens for evidence or development. The tutor listens for language and interaction.

Repair

If the listener misunderstands, the speaker must reformulate rather than repeat the same sentence louder.

This links to Speaking & Oral Communication.

Pronunciation

Pronunciation correction should target intelligibility, not accent ranking.

Three students can test whether the same word is recoverable to different listeners. If all three mishear the same contrast, the pronunciation target is clearer. If only one unfamiliar listener struggles, exposure may be part of the issue.

World variation is covered by World Englishes & Variation.

Oral discussion

Three learners create a natural minimum discussion group. Teach:

• entry into discussion;
• building on a previous point;
• asking clarification;
• disagreement with reasons;
• inviting a quieter speaker;
• synthesising where the group has reached.

The tutor should avoid becoming the permanent fourth participant.

Examination performance: decisions under constraint

Examination training needs more than content knowledge. It needs observable decisions under time.

Timed start

All three begin the same section together. The tutor records milestones rather than interrupting.

Post-task reconstruction

After time, each student marks:

• where they slowed;
• what they skipped;
• what they changed;
• what they did not finish;
• which decision they would repeat or change.

Compare strategy with outcome

A fast student may have low accuracy. A slow student may be overchecking. A student with the best score may still use a fragile strategy that will fail on a harder paper.

Performance analysis therefore examines both result and method.

Recovery training

Give each student a deliberately difficult item. The target is not solving it perfectly. The target is containment.

Practise:

• recognising diminishing returns;
• marking the item;
• moving on;
• resetting attention;
• returning with a defined method;
• submitting a complete paper.

Leonie’s role as an execution/recovery example is especially useful here.

Primary, Secondary and JC use the same model differently

Primary

Shorter cycles, more oral explanation, concrete examples, visible routines and parent-readable feedback. The tutor protects curiosity while building foundational control.

Secondary

Greater independence, denser texts, broader writing forms, peer comparison of reasoning and explicit examination decision-making.

JC / advanced English

Longer reading, competing positions, evidence weighting, argument synthesis, timed essay architecture and spoken defence of reasoning.

The class size stays three. The cognitive work changes dramatically.

How to know whether a three-student tutorial is actually working

Class size is easy to observe. Learning is harder. A family can see that there are only three students and still have very little evidence about whether the tutorial is improving independent English performance.

SETC therefore separates format evidence from learning evidence.

Format evidence includes:

• how many students are present;
• how often each learner receives direct attention;
• whether independent work happens;
• whether peer comparison is purposeful;
• whether feedback is student-specific.

Learning evidence includes:

• what the learner can now do that they could not do before;
• whether the improvement survives new material;
• whether fewer prompts are needed;
• whether time and accuracy improve under realistic conditions;
• whether the learner can explain and self-correct the process;
• whether school or examination performance becomes more stable for reasons that can be traced to the trained skill.

The group size can create the conditions. Only the second list shows whether those conditions are being used well.

Baseline before improvement

Progress claims are stronger when the starting state is visible.

A baseline can be small:

• one unseen reading passage;
• one writing sample;
• one oral response;
• one timed section;
• one vocabulary retrieval task;
• one grammar/editing sample.

The baseline should match the intended goal. If tuition is supposed to improve writing, a vocabulary quiz alone is not an adequate baseline.

Record process evidence

Where possible, record not only the score but the failure pattern.

Example:

Comprehension: 6/10. Two marks lost through unsupported inference, one through excessive answer scope, one through incomplete paraphrase.

This gives later improvement a mechanism.

Short-cycle evidence

Not every improvement needs to wait for an examination. Tutorials can collect short-cycle evidence.

Reading

Fewer unsupported inferences across three fresh passages.

Writing

Paragraph plans show clearer jobs before drafting; revisions require fewer tutor prompts.

Grammar

The same sentence relationship is controlled correctly across unfamiliar examples.

Vocabulary

Target words are retrieved and used naturally after delay.

Listening

One missed detail no longer disrupts later answers.

Speaking

The learner develops a response without a tutor supplying the second reason.

Examination performance

Time checkpoints become more stable; fewer blanks remain; checking produces fewer harmful answer changes.

Delayed evidence

Immediate success after feedback is weak evidence because the learner still remembers the explanation.

Retest later with:

• different topic;
• different wording;
• different text;
• different order;
• less prompting;
• longer delay.

Delayed success increases confidence that the learner has internalised something.

School marks as evidence

School marks matter, but they are noisy. Paper difficulty, topic familiarity, marking, fatigue and cohort performance can change.

Use marks alongside mechanism evidence.

If writing marks rise and the learner’s planning, paragraph control and revision independence also improve, the story is stronger than “marks went up after tuition”.

When marks do not rise immediately

A learner may be repairing a deep weakness while current school work still exposes older gaps. Improvement can also be hidden when the paper becomes harder.

Do not use this as an excuse indefinitely. The tutor should still be able to show intermediate evidence:

• better task interpretation;
• more accurate evidence use;
• fewer recurring errors;
• stronger independent drafts;
• more stable timing.

When marks rise for the wrong reason

A learner can improve on familiar worksheets or memorised model structures without gaining transfer.

Check whether the improvement survives:

• unseen questions;
• new text types;
• different topics;
• reduced tutor support;
• time pressure.

If performance collapses, the training may be too narrow.

Progress records should stay interpretable

A tutor does not need a giant dashboard. A useful record can track:

The record tells a learning story rather than accumulating scores without explanation.

Parent communication

Parents need information they can use. “She participated well” is pleasant but weak evidence.

A stronger update might say:

This week we tested evidence selection in comprehension. She initially relied on repeated keywords, which caused two unsupported answers. After practising a meaning-first test, she selected the correct evidence on three new questions. Next week we will test the same decision in an unfamiliar passage without the checklist.

The parent knows:

• what was trained;
• what failed;
• what changed;
• what remains to be proven.

Do not report every mistake to parents

Constant error reporting can make English learning feel like surveillance. Parent communication should focus on patterns, progress, important risks and useful support at home.

What parents can support

Parents do not need to reproduce the tutorial.

Useful support can include:

• reading time;
• conversation about ideas;
• encouraging the learner to explain a choice;
• protecting homework routines;
• asking what the learner is practising independently;
• reducing last-minute rescue.

The Parent Support route covers this more fully.

Group composition matters

Three students are not interchangeable units. A tutorial group has dynamics.

Age and developmental stage

A Primary 2 learner and Secondary 2 learner should not share the same ordinary English tutorial simply because both “need grammar”. Their tasks, language demands and attention patterns differ.

Curriculum and examination demands

Students can sometimes benefit from mixed schools or syllabuses when the mechanism is shared. But if the lesson is paper-specific, major format differences can reduce useful shared time.

Current ability

Perfectly matching scores is unnecessary. Some difference can enrich contrast. Extremely large gaps can create two lessons happening at once.

Learning state

One learner may need foundational repair while another needs advanced extension. If the common task cannot expose useful decisions for both, the grouping is weak.

Behaviour and participation

One highly dominant learner can suppress two others. One learner requiring constant behavioural regulation can consume the observation budget. Grouping should account for the learning environment, not merely timetable convenience.

Friendship

Friends can create safety and motivation or distract one another. The tutor needs evidence from actual behaviour rather than assuming friendship is good or bad.

Should all three students be at the same level?

No absolute rule. A small difference can support peer contrast. The tutor can assign different complexity while preserving a common mechanism.

For example, all three students practise evidence-supported inference, but texts differ slightly in density. Discussion can still focus on the shared method.

The group stops working when differentiation becomes so large that the tutor is effectively running three unrelated lessons.

When the three-student format may not be the best fit

SETC should say this clearly because a teaching model becomes more credible when its boundaries are visible.

Highly private or sensitive learning needs

A learner may need one-to-one support where privacy is essential.

Severe foundational mismatch

If the learner cannot access the common task at all, repeated differentiation may create frustration.

Intensive short-term specialist preparation

Some highly specialised objectives may benefit from one-to-one work for a period, especially when every minute must target one learner’s unique task.

Group dynamics that repeatedly block learning

If two learners continually distract one another or one learner cannot participate safely, changing the group can be better than forcing the model.

Schedule instability

A three-student model depends on enough continuity for shared routines and transfer checks. Constantly changing attendance can weaken it.

The learner needs a different professional service

Not every difficulty is a tuition problem. Hearing, vision, speech-language, psychological, medical or other needs may require appropriate qualified professionals. English tutoring should stay inside its competence.

When one-to-one may be better temporarily

A learner can move between formats over time. One-to-one support may be useful for:

• initial diagnostic work;
• acute examination recovery;
• highly specialised writing;
• private confidence rebuilding;
• bridging a large foundational gap before joining a group.

The later question is whether the learner can benefit from peer contrast and more independent performance.

When a larger group may be better

Larger groups can be excellent for:

• debate;
• literature discussion;
• project work;
• large-scale collaborative tasks;
• explicit teaching where constant individual diagnosis is unnecessary;
• activities where a wider diversity of viewpoints is itself the objective.

The three-student model should not claim every educational objective.

Online three-student tutorials

Online delivery can preserve many advantages if the technology and task design support visibility.

Useful online features

• shared documents with individual cursors;
• screen sharing;
• private submission before group discussion;
• recorded oral responses;
• breakout only when needed;
• shared annotation.

Online risks

• camera-off disengagement;
• hidden use of external tools during independent work;
• technical delays;
• reduced access to handwriting or physical working;
• digital fatigue.

Online teaching should design around those risks rather than pretend the medium is identical.

In-person three-student tutorials

In-person work gives the tutor access to physical pages, eye movement, hesitation, body orientation and rapid informal interaction. It can also create travel cost and scheduling constraints.

The better medium depends on learner, task and logistics.

Hybrid teaching

A group can combine formats across a term: in-person writing workshops, online review, recorded oral tasks and asynchronous transfer practice. The teaching model remains the same if diagnostic visibility and independent evidence remain strong.

Use of AI in a three-student tutorial

AI can generate additional examples, simulate audience responses or offer provisional feedback. It can also erase the diagnostic evidence by producing the answer.

Use AI where the learner must still make the target decision.

Examples:

• compare two AI-generated explanations and identify which is more accurate;
• find an unsupported claim in an AI paragraph;
• rewrite an AI answer for a different audience;
• verify an AI vocabulary suggestion through dictionaries/corpora;
• complete an unaided task after AI-supported practice.

The Digital & AI English owner provides the wider framework.

AI should not become the fourth student

The group’s educational value comes from three human decision processes becoming visible. If every uncertainty is immediately sent to an AI system, the students stop learning to evaluate one another and themselves.

Use the tool deliberately, not continuously.

Matching tutors to the model

A three-student tutorial requires different skills from lecturing to a class.

The tutor needs to:

• observe process;
• ask diagnostic questions;
• manage turn-taking;
• differentiate without splitting the room into three unrelated lessons;
• give concise feedback;
• know when not to intervene;
• fade support;
• track transfer.

Subject knowledge remains essential. The format does not compensate for weak English expertise.

What a parent should ask before choosing a three-student class

1. How are students grouped?
2. How does the tutor diagnose individual weaknesses?
3. What happens when students have different errors?
4. How much independent work occurs during the lesson?
5. How is feedback recorded?
6. How is progress retested on unseen tasks?
7. How do you prevent one student from dominating?
8. What happens if the group fit is poor?
9. When would you recommend one-to-one instead?
10. How do you know when the student no longer needs the same level of support?

These questions reveal the teaching system behind the marketing label.

Red flags

A “three-student class” deserves scrutiny if:

• the tutor lectures for most of the lesson;
• students copy notes with little independent production;
• everyone receives identical feedback despite different errors;
• homework is never used diagnostically;
• model answers are memorised;
• the tutor supplies answers before students commit to a choice;
• progress is described only through attendance or worksheet completion;
• students become less willing to work without prompts;
• the group remains together despite obvious incompatibility;
• parents are told that three students is automatically superior because it is smaller.

Green flags

• students attempt independently before comparison;
• the tutor can name the current weak link for each learner;
• peer discussion uses criteria;
• feedback leads to a specific retry;
• the same mechanism is tested on new material;
• support reduces as competence grows;
• students can explain their own errors;
• the tutor can say when the format is not the best fit.

The fit test

The three-student model is working when the common task creates useful individual evidence, the peer contrast improves judgement, tutor attention remains sufficiently precise, and each learner becomes more independent.

If any of those four conditions repeatedly fails, the format or grouping needs to change.

Worked tutorial casebook: three learners, different weak links

The examples below are instructional simulations. Maren, Iona and Leonie are stable resident students used to make decisions visible. They are not claims about identifiable children or promises of outcomes.

Case 1: Primary reading — same wrong answer, different reasons

The passage says:

Amir folded the letter twice, placed it under the book and looked toward the doorway whenever footsteps passed.

The question asks: What suggests that Amir does not want someone to see the letter?

All three students initially write an incomplete answer.

Iona

Iona selects “looked toward the doorway” because it clearly indicates concern about someone arriving. She misses the hidden-letter evidence.

The tutor asks: “Which action changes the visibility of the letter?”

Iona adds “placed it under the book”. Her weak link is evidence completeness, not inference.

Maren

Maren selects “folded the letter twice” and explains that folding means secrecy.

The tutor asks: “Could someone fold a letter for another reason?”

Maren realises the inference is possible but weak. Her weak link is evidence strength.

Leonie

Leonie initially selects both strong pieces of evidence, then deletes one because she worries the answer is too long.

Her weak link is not reading. It is checking behaviour.

One question. Three repairs.

Case 2: Primary vocabulary — knowing the word but not owning it

The target word is reluctant.

Iona defines it accurately as not wanting to do something.

Maren gives the sentence: “She was reluctant because the weather was reluctant.” The definition is known, but grammatical and semantic use is unstable.

Leonie writes a natural sentence but cannot retrieve the word two days later.

The tutorial distinguishes:

• meaning;
• sentence behaviour;
• collocation;
• retrieval strength.

A conventional vocabulary quiz might treat all three students similarly. The three-student comparison reveals different states.

Case 3: Primary composition — the exciting idea that cannot carry a story

The prompt is an image of a damaged bicycle near a playground.

Maren proposes a dramatic theft story with four characters, a chase and a surprise police arrival.

Iona proposes that the bicycle was damaged because a friend borrowed it without permission and tried to hide what happened.

Leonie proposes a simple accident followed by a decision about whether to tell the owner.

The tutor asks each to identify:

• character goal;
• decision;
• consequence;
• how the story can end causally.

Maren discovers that her idea contains excitement but no controllable decision chain. Iona has strong conflict but needs an ending. Leonie has the clearest causal structure but needs richer scene development.

The tutorial does not choose one “best plot”. It helps each student strengthen the architecture they selected.

Case 4: Primary oral — prepared answer vs responsive control

The students prepare for the question: Should children have more time to play outdoors?

Maren gives a fluent two-reason answer.

Iona gives a shorter answer with one clear personal example.

Leonie hesitates but explains one thoughtful trade-off between outdoor play and homework.

The tutor then asks an unseen follow-up: What if the weather is unsafe or there is no nearby outdoor space?

Maren repeats her prepared points. Iona adapts and proposes indoor physical activities. Leonie reframes her answer around access and safety.

The follow-up reveals responsiveness that the prepared answer did not.

Case 5: Secondary comprehension — keyword matching

A question asks why the writer describes a new policy as “a bridge rather than a destination”.

Iona searches for another sentence containing bridge. None exists.

Maren interprets the metaphor but gives a broad answer about change.

Leonie explains that the policy is presented as a temporary or transitional step rather than the final outcome.

The tutor asks Iona to stop keyword hunting and identify the relationship inside the metaphor. Maren is asked to specify the contrast. Leonie is asked what evidence would make the interpretation unsafe.

Three students work at three levels of the same inference.

Case 6: Secondary summary — too much truth

All three understand the source passage.

Iona includes every accurate detail and exceeds the word limit.

Maren removes so much that one central relationship disappears.

Leonie selects the right points but repeats one idea in two forms.

The tutor uses a proposition table:

Students learn selection as a hierarchy rather than word cutting.

Case 7: Secondary situational writing — right format, wrong reader

The task asks students to write an email to a school leader proposing a change to a school event.

Maren uses the correct email format and formal vocabulary but spends two paragraphs describing her personal feelings.

Iona includes every source detail but never states the requested action clearly.

Leonie states the action early but sounds overly blunt.

The three peer roles are:

• reader need;
• source completeness;
• tone/relationship.

Each student receives a different revision while the common communication system remains visible.

Case 8: Secondary argumentative writing — evidence without warrant

The claim is: Schools should limit smartphone use during lessons.

Maren provides a study showing students receive many notifications.

Iona provides an example of one distracted student.

Leonie argues that distraction can reduce sustained attention but gives no evidence.

The tutor separates three needs:

• Maren must explain why notification frequency matters to learning.
• Iona must avoid treating one anecdote as prevalence evidence.
• Leonie needs support for the causal or practical claim.

This links rhetoric, reasoning and evidence.

Case 9: Secondary grammar — pronoun reference

Sentence:

When the teachers met the parents after they received the results, they were concerned.

Who received the results? Who was concerned?

The ambiguity is real.

Each student rewrites the sentence differently:

• name the actor;
• split the sentence;
• change clause order.

The tutor compares which rewrite is clearest for the intended meaning.

Grammar becomes reader control, not rule recitation.

Case 10: Secondary listening — the missed number

A recording gives several dates and times.

Iona hears the wrong date but continues listening and answers later questions correctly.

Maren records every number and loses the explanation around them.

Leonie misses one number, panics and stops processing the next sentence.

The tutorial target becomes different:

• Iona: verification of high-risk information;
• Maren: selective note-taking;
• Leonie: recovery after uncertainty.

Case 11: Secondary examination timing — equal marks, unequal risk

All three score 18/25 on a timed section.

Iona finishes with five minutes left but loses marks through rushed reading.

Maren uses every minute and leaves one item blank.

Leonie finishes all items but changes two correct answers during final checking.

The same score does not justify the same timing advice.

Case 12: JC GP reading — source position vs student opinion

A passage argues that technological convenience can reduce deliberate human judgement.

Maren agrees with the writer and begins inserting her own examples before accurately summarising the source.

Iona represents the source precisely but struggles to evaluate its assumptions.

Leonie identifies a limitation but overstates it as a complete refutation.

The tutor assigns:

• Maren: source representation before response;
• Iona: assumption testing;
• Leonie: proportional evaluation.

Case 13: JC GP essay — knowledge dump

The question concerns whether cities should prioritise efficiency over community.

Iona has six memorised examples and tries to use four.

Maren has two examples but a strong conceptual distinction between efficiency and social resilience.

Leonie has a balanced thesis but cannot retrieve enough specific evidence quickly.

The three-student comparison teaches that knowledge quantity, conceptual quality and retrieval speed are different variables.

The tutor asks each student to build one paragraph:

• claim;
• evidence;
• mechanism;
• qualification;
• return to question.

Case 14: JC GP synthesis — flattening disagreement

Two authors both support regulation but for different reasons.

Iona writes that “both authors agree regulation is necessary”. Accurate but shallow.

Maren identifies that one author emphasises consumer protection while the other emphasises market stability.

Leonie notices that their preferred regulatory mechanisms differ.

The group builds a synthesis ladder:

1. surface agreement;
2. reason;
3. mechanism;
4. scope;
5. implication.

Each learner then writes independently.

Case 15: vocabulary at advanced level — impressive but imprecise

Maren writes ubiquitous when she means merely common.

Iona uses ameliorate correctly but in a context where improve is clearer.

Leonie avoids advanced vocabulary entirely because she fears sounding unnatural.

The tutor uses Reference Literacy to compare frequency, collocation and register.

The lesson is lexical precision, not sophistication display.

Case 16: academic source use — three kinds of paraphrase failure

The source says that a feedback method may improve immediate revision accuracy under structured conditions.

Iona copies the sentence structure and changes a few words.

Maren rewrites independently but changes may improve to improves.

Leonie preserves meaning but removes under structured conditions.

Three failures:

• patchwriting;
• claim-strength distortion;
• scope distortion.

The group learns why paraphrase is meaning preservation, not synonym replacement.

Case 17: rhetoric — persuasive power without evidence

Students evaluate a short advertisement.

Maren notices emotional language.

Iona notices a testimonial.

Leonie notices the advertisement never supplies evidence for the central performance claim.

The tutor connects the three observations:

emotion + borrowed credibility + missing evidence.

Rhetorical analysis becomes architecture.

Case 18: AI-assisted writing — strong product, weak learner evidence

All three students use an AI assistant during practice.

Maren asks it for a full rewrite and submits polished prose.

Iona asks it to identify two weak sentences, then revises them herself.

Leonie asks for an opposing argument and writes her own response.

The final products may all improve. The learning evidence differs.

The tutor removes AI and gives a matched unseen task. That task determines which support produced transferable capability.

Case 19: one student improves faster

After six weeks, Iona no longer needs the group’s reading-support routine. Maren and Leonie still benefit.

The tutor should not hold Iona artificially at the same level. Extension can shift her role:

• evaluate competing interpretations;
• work with denser texts;
• justify why an answer is insufficient;
• transfer the same reading mechanism into GP-style sources.

Group teaching can differentiate vertically while sharing the core mechanism.

Case 20: one student needs a different environment

A student repeatedly becomes distressed during timed comparison, requires privacy for feedback and cannot participate safely in peer contrast.

The teaching model should yield to the learner’s need. One-to-one or another support structure may be more appropriate.

A model that cannot admit its own boundary is not a serious model.

What the casebook teaches

Across all twenty cases, three principles recur:

1. The final error is not the diagnosis.
2. Shared tasks can produce different individual repairs.
3. The repair is not complete until it survives independent transfer.

That is the educational logic behind three students.

The tutor’s diagnostic system: observe the learner without turning observation into a label

A three-student tutorial gives the tutor unusually rich information. That creates a responsibility: information should improve teaching, not produce premature stories about the learner.

“Careless”, “weak”, “lazy”, “not a language person” and “overthinker” are low-resolution labels. They compress different mechanisms into personality.

A diagnostic tutor asks instead:

• What task was attempted?
• What did the learner understand?
• Which decision failed first?
• Under what conditions does the problem appear?
• Can the learner perform with a prompt?
• Can the learner perform after the prompt is removed?
• Does the weakness repeat across tasks?
• What evidence would prove the repair?

Separate knowledge from performance

A student can know something and fail to retrieve it under time. Another can perform correctly through a memorised template without understanding the underlying relationship.

Three students give useful contrasts:

Iona explains the grammar rule accurately but cannot apply it while writing quickly.

Maren applies the structure correctly in a familiar exercise but cannot explain when it should be used.

Leonie produces the correct form and explanation but checks so often that the task is unfinished.

Knowledge, transfer and execution are separate diagnostic dimensions.

A five-layer diagnostic model

Layer 1: access

Can the learner decode or understand the task, vocabulary and instructions?

If access fails, later reasoning cannot be interpreted cleanly.

Layer 2: representation

Has the learner built the correct model?

For reading: what does the passage say?

For writing: what does the question require?

For grammar: what relationship does the sentence express?

Layer 3: decision

Can the learner choose the relevant evidence, structure, word, grammar form or strategy?

Layer 4: execution

Can the learner carry out the decision accurately enough and within the available time?

Layer 5: monitoring and recovery

Can the learner detect problems, repair them and continue without destroying the rest of the task?

The first weak layer often determines the intervention.

Do not diagnose motivation before diagnosing the task

A student who avoids writing may be unmotivated. They may also have no method for generating ideas, fear repeated correction, lack vocabulary for the topic or believe every sentence must be perfect before continuing.

Ask for process evidence before attributing the behaviour to attitude.

Do not diagnose confidence as a single trait

A learner can be confident in oral discussion and hesitant in writing. They can be confident on familiar comprehension and uncertain on inference.

Confidence is often task-specific and evidence-responsive.

Do not diagnose “carelessness” without locating the error timing

A so-called careless error can occur because:

• the learner never knew the rule;
• the learner knew it but attention was overloaded;
• the learner changed a correct answer while checking;
• the learner rushed because earlier tasks consumed time;
• the learner misread one word in the question;
• the learner uses a weak checking routine.

Each requires a different repair.

Diagnostic questions should expose a model, not lead to the answer

Compare:

“Don’t you think this pronoun refers to the teacher?”

with:

“What are the possible referents? What grammatical and meaning evidence supports each?”

The first question teaches guessing the tutor’s intention. The second exposes reasoning.

Use contrast cases

A contrast case changes one important feature while holding others stable.

If a learner thinks although and because are interchangeable connectors, give two sentences where only the logical relationship changes.

If a learner overgeneralises a vocabulary word, give one natural and one unnatural context.

Contrast sharpens category boundaries.

Use counterexamples

When a student proposes a rule, test it.

Student rule: “A conclusion should always introduce no new idea.”

Tutor asks: “Can a conclusion state an implication that follows from the argument even if those exact words did not appear earlier?”

The goal is not to destroy rules but refine them.

Use prediction

Ask learners to predict before revealing an answer or model.

Prediction creates visible commitment. After feedback, the learner can compare the mental model with evidence.

Use explanation after correctness

A correct answer does not prove a stable method. Ask:

“What made that answer stronger than the other option?”

If the learner cannot explain, success may be fragile.

Use transfer after explanation

An excellent explanation does not prove the learner can perform. Give a new task.

The diagnostic chain becomes:

answer → explanation → transfer.

The three-student diagnostic triangle

The tutor can compare across three dimensions:

• within learner: how this student changes across tasks and time;
• between learners: how different strategies affect the same task;
• against task criteria: what successful performance actually requires.

No one comparison is sufficient alone.

Do not rank students publicly

Peer contrast should not become a permanent hierarchy: strongest, middle, weakest.

Ranking reduces complex profiles to one dimension and can distort group behaviour.

Use local language:

“Iona’s evidence choice is strongest on this question.”

not:

“Iona is the best reader.”

Diagnostic notebooks

A tutor can maintain compact notes for each learner:

The notebook should drive teaching. If notes are never used to choose tasks, they are bureaucracy.

Group-level notes

Some patterns belong to the group:

• all three misread a command word;
• all three have weak knowledge of one genre;
• discussion repeatedly becomes dominated by one student;
• homework completion is inconsistent;
• the current material is too easy to expose useful differences.

Group-level problems require group-level changes.

From diagnosis to lesson sequence

One weak link should create a sequence, not one isolated correction.

Example: evidence selection

Lesson 1: diagnose evidence choice using short questions.

Lesson 2: contrast strong and weak evidence on different text types.

Lesson 3: integrate evidence choice into full comprehension.

Lesson 4: timed unseen task.

Lesson 5: delayed retest without reminder.

The skill is not “completed” after Lesson 1.

Spiral back without teaching the same lesson again

A concept can return under greater complexity.

Evidence selection appears in:

• Primary comprehension;
• Secondary inference;
• argumentative writing;
• GP source evaluation;
• academic research.

The mechanism spirals through new demands.

Interleaving

Once several skills are stable enough, mix them. A student should decide whether the next problem requires inference, vocabulary, grammar, source evaluation or timing.

Interleaving is especially useful for examinations because the paper does not announce the cognitive mechanism before every question.

Retrieval

Ask learners to reconstruct the method before displaying the checklist.

“What are the three things you check before answering an inference question?”

If the learner can retrieve the process, the scaffold is moving inward.

Spacing

Return after enough time that the answer is not held in working memory. Spacing increases the diagnostic value of success.

Variation

Change surface features so the learner must recognise the underlying structure.

For writing, change topic and audience.

For grammar, change vocabulary while preserving the relationship.

For reading, change genre.

Near transfer

Near transfer changes the surface modestly. Example: same comprehension mechanism in a different passage.

Farther transfer

Farther transfer requires applying the principle in a different context. Example: evidence discipline learned in comprehension supporting source use in argumentative writing.

Failure after transfer is useful evidence

If a learner succeeds in drills and fails in integration, the tutor has learned something important: the skill is not yet accessible under divided attention.

Return to recombination rather than repeating isolated drills indefinitely.

Progression criteria

Move a learner forward when:

• the target decision is accurate across varied examples;
• the learner can explain it;
• performance survives reduced prompting;
• the learner can integrate it with neighbouring skills;
• delayed transfer is acceptable.

Do not progress merely because a worksheet was finished.

Regression and forgetting

Performance can decline after a break or under new pressure. This does not necessarily mean the learning vanished.

Test how quickly it can be reactivated. A previously learned skill that returns after one reminder is in a different state from a skill that must be rebuilt from zero.

When to stop teaching a target explicitly

Explicit teaching should reduce when the learner:

• retrieves the rule or method independently;
• detects common errors;
• uses the skill in integration;
• recovers after mistakes;
• transfers across new tasks.

Continue monitoring without occupying lesson time unnecessarily.

Tutor calibration

Teaching quality depends on the tutor’s ability to update their diagnosis.

A hypothesis is not an identity:

“I think the current bottleneck is evidence selection.”

Test it.

If evidence selection is accurate on several fresh tasks, update the hypothesis.

False diagnosis from one bad day

Fatigue, stress, illness, unfamiliar topic and distraction can distort performance. Look for patterns across evidence before redesigning the entire curriculum.

False diagnosis from one good day

Topic familiarity, lucky guessing or recent rehearsal can inflate performance. Verify across variation and delay.

The tutor’s stop rule

Do not keep diagnosing indefinitely. Once enough evidence identifies a useful repair, teach it.

Diagnosis serves action.

Diagnostic humility

English tutors observe learning performance; they are not automatically qualified to diagnose medical, psychological or developmental conditions. When evidence suggests a need outside English teaching, the correct action is to recommend appropriate professional assessment rather than invent a label.

What three students uniquely contribute to diagnosis

The group does not give the tutor three times as much truth. It gives three simultaneous comparison cases under a shared task.

That can help the tutor notice:

• which errors are task-wide;
• which are student-specific;
• which explanations work for more than one learner;
• which strategies transfer;
• which student requires a different task.

The format turns contrast into instructional evidence.

The diagnostic promise SETC should make

Not:

“With only three students, every child improves faster.”

But:

“With three students, the teaching can be designed so individual English decisions remain visible enough to diagnose, compare, repair and retest—while students still learn from seeing other minds solve the same problem.”

That claim is narrower, more defensible and more useful.

Independence is the exit condition, not an optional extra

A tuition system can produce excellent-looking work while increasing dependence. The learner waits for the tutor to interpret the question, approve the plan, choose the evidence, correct the sentence and decide whether the work is finished.

The final product improves. The learner’s decision system does not.

SETC treats independence as a design requirement. Every recurring tutor action should eventually face a question:

Can the learner take over this action?

From external questions to internal questions

Early in a reading repair, the tutor may ask:

What exactly is the question asking?

Later, the student should ask it silently before answering.

Early in writing:

What is this paragraph’s job?

Later, the student should generate the job before drafting.

Early in examination work:

Are you spending too long on this item?

Later, the learner should notice diminishing returns and move on.

Teaching succeeds when tutor prompts become learner routines.

The support ladder

Support can be represented as a ladder from most explicit to least explicit:

1. tutor models the complete process;
2. tutor models and explains decisions;
3. learner completes part of the process with prompts;
4. learner completes the process with a checklist;
5. learner completes it with only a general reminder;
6. learner initiates the process independently;
7. learner adapts the process to a new task;
8. learner diagnoses failure and chooses a repair.

A learner may occupy different ladder positions in different skills.

Do not remove support by calendar

“After four weeks, remove the checklist” is weaker than “remove the checklist after the learner can reconstruct and apply it reliably across varied examples.”

Fading should follow evidence.

Partial fading

If full removal causes collapse, restore only the smallest useful support.

Instead of returning the entire paragraph planner, give the question:

What must the reader understand by the end of this paragraph?

The goal is not to prove the learner can fail independently. It is to find the minimum support that restarts productive performance.

Self-explanation

Learners gain independence when they can explain their own choices.

Useful prompts:

• Why did you choose this evidence?
• What alternative did you reject?
• What makes this word more precise?
• Why did you stop checking here?
• Which rule are you applying?
• How would the answer change if the audience changed?

Self-explanation exposes whether the method is owned or merely imitated.

Self-questioning

Students can build small internal checklists.

Reading self-questions

• What is the question job?
• What is the strongest evidence?
• What must I infer?
• What is outside scope?

Writing self-questions

• What am I trying to make the reader understand?
• What does this paragraph contribute?
• Does the evidence support the claim?
• What should I revise first?

Oral self-questions

• Did I answer the exact question?
• Have I developed the point?
• Is the listener following?
• Can I reformulate?

Exam self-questions

• Am I still gaining marks from the time I am spending?
• What should I skip and return to?
• What needs checking most?
• Is one difficult item controlling the rest of the paper?

Metacognition without endless reflection

Reflection is useful when it changes the next action. A ten-minute diary after every worksheet can become ritual.

Keep reflection short:

What worked? What failed? What will I do differently next time?

Learning to diagnose your own error

The tutor can gradually ask students to classify errors.

Example categories:

• question interpretation;
• knowledge gap;
• evidence selection;
• reasoning;
• language form;
• timing;
• checking;
• recovery.

A learner who can identify the category can choose a better repair.

Error logs

An error log should track patterns, not become a museum of every mistake.

Self-marking

Self-marking is useful after independent commitment. If students see the model first, they practise recognition.

Use:

1. attempt;
2. mark independently;
3. compare with criteria/model;
4. explain discrepancy;
5. revise;
6. retest on new item.

Peer marking

Peer marking can help when criteria are narrow. It becomes unreliable when students judge broad quality without enough expertise.

Use peers for observable features:

• Is evidence present?
• Is the main claim identifiable?
• Does every pronoun have a clear referent?
• Did the response answer all parts?

Keep expert judgement where expert judgement is needed.

Choice of strategy

Independence does not mean using one strategy automatically. It means selecting among strategies.

A strong reader decides when to skim, scan, reread, annotate or stop.

A strong writer decides when to outline extensively and when a short plan is enough.

A strong examiner decides when to persist and when to defer.

Strategy choice is a higher stage than strategy compliance.

Strategy cost

Every strategy consumes time and attention. A ten-step checklist can improve accuracy during training and become too costly under examination conditions.

As learners improve, compress routines.

“Question → evidence → scope” may replace a longer reading checklist.

Automaticity

Some lower-level operations should become faster so attention can move upward.

Examples:

• common word recognition;
• basic punctuation;
• frequent grammatical patterns;
• routine paragraph transitions;
• common examination instructions.

Automaticity does not eliminate thinking. It frees thinking capacity.

Deliberate control remains necessary

Advanced tasks still require slow judgement. A GP student should not “automate” evaluation in the same way they automate spelling.

The tutor helps learners distinguish which operations should become automatic and which should remain deliberative.

The independence paradox

Students often become more independent after receiving well-designed support, not after being left alone prematurely.

Support and independence are not opposites. Bad support replaces thinking. Good support makes future thinking possible.

Transfer dimensions

A skill is stronger when it survives changes in:

• topic;
• genre;
• difficulty;
• surface vocabulary;
• time;
• medium;
• audience;
• support;
• emotional pressure;
• combination with other skills.

Fresh-task transfer

Use material the student has not practised. If the tutor helped build the exact answer earlier, the task is review, not transfer.

Delayed transfer

Wait. A method available only immediately after instruction may not be retrievable later.

Cross-skill transfer

Evidence discipline learned in reading can support writing.

Audience awareness learned in situational writing can support oral presentation.

Reference literacy learned in vocabulary can support Academic English.

Cross-skill transfer is one of the strongest signs that learners understand the underlying principle.

Cross-stage transfer

Primary evidence selection becomes Secondary inference discipline. Secondary argumentative structure becomes GP reasoning. GP source use becomes Academic English.

SETC’s Atlas makes those longitudinal relationships explicit.

Transfer failure types

Surface overfitting

The learner succeeds only when examples look like training examples.

Prompt dependence

The learner succeeds only when a checklist or tutor question is present.

Context dependence

The learner succeeds in tuition but not school because task conditions differ.

Time collapse

The learner can perform untimed but not under examination constraints.

Integration collapse

The learner can perform the skill alone but not when combined with reading, writing and timing demands.

Retrieval failure

The learner understood previously but cannot access the method after delay.

Each failure suggests a different next practice.

Transfer design

A good sequence deliberately increases distance.

Example for inference:

1. short familiar passage with prompt;
2. short new passage without prompt;
3. longer passage;
4. different genre;
5. timed section;
6. inference embedded inside synthesis;
7. delayed unseen assessment.

Release tests

A release test asks whether the learner still needs the tutorial intervention.

Reading release

Can the learner interpret questions, select evidence and control inference across unfamiliar texts without the tutor’s cue?

Writing release

Can the learner interpret a prompt, structure, draft, revise and edit without the tutor deciding what to write?

Grammar release

Can the learner detect and repair common target errors inside real writing?

Vocabulary release

Can the learner learn new words independently through context, dictionaries, corpora and retrieval?

Oral release

Can the learner respond, develop, clarify and repair under unseen questioning?

Examination release

Can the learner manage timing, uncertainty and recovery without external prompting?

Leaving a specific repair

The learner may continue tuition while one target is released. Do not keep practising a stable skill because it is comfortable to teach.

Move instructional time to the next limiting factor.

Reducing lesson frequency

As independence grows, some learners may benefit from wider intervals so they accumulate real independent work between sessions.

Spacing can turn the next tutorial into an audit and refinement session rather than continuous supervision.

Ending tuition

A responsible tutorial model includes the possibility that tuition is no longer necessary for the current objective.

Possible signs:

• school and unseen performance are stable;
• the learner self-diagnoses common issues;
• new material is learned independently;
• feedback is converted into future changes;
• the tutor’s prompts rarely add value;
• the learner has a workable independent study system.

Ending or pausing tuition can be evidence of success.

Changing the objective

A learner may no longer need repair but choose extension: literature, rhetoric, Academic English, advanced writing, GP or broader intellectual work.

The relationship changes from remediation to development.

Dependency warning signs

• student will not start until tutor approves the plan;
• student asks “Is this right?” after every sentence;
• student performs far better in tuition than on unseen school work;
• student waits for model answers before attempting;
• student cannot explain corrections;
• student’s work becomes polished while speed and independence decline.

These signals require scaffold reduction.

Tutor withdrawal as an instructional skill

Some teaching expertise lies in knowing what not to say.

A tutor can deliberately:

• pause before answering;
• redirect a question back to evidence;
• ask another student for a criterion;
• remove a checklist;
• withhold confirmation until the learner commits;
• let a safe error play out so the learner can diagnose it.

Students learning to help one another without dependence

Peer support should also fade from answer-giving toward question-giving.

Instead of:

“The answer is paragraph three.”

use:

“Which sentence actually proves the change happened?”

The peer preserves the other learner’s decision.

Independence is compatible with collaboration

Independent learners can still discuss, ask for feedback, use references and collaborate. Independence means they use support strategically rather than being unable to function without it.

Advanced independence

An advanced learner can:

• recognise when a task exceeds current knowledge;
• choose a reliable reference;
• seek targeted feedback;
• evaluate the feedback;
• integrate useful changes;
• reject inappropriate advice with reasons;
• test their own performance.

This is not isolation. It is self-regulation.

The tutorial’s final transfer loop

Learn → practise → vary → integrate → delay → perform unseen → diagnose independently → seek help only where needed.

The three-student environment can support every stage, but it should not become a permanent substitute for the final unseen performance.

The successful disappearance of the tutor

A strong teacher leaves traces in the learner’s decisions.

The student may no longer hear the tutor’s voice explicitly, but the questions remain:

What does the task ask?

What evidence supports this?

What is my reader likely to understand?

What changed after the feedback?

Can I do this on a fresh task?

When those questions become internal, the tutorial has done something more valuable than produce a good worksheet.

What the wider evidence does—and does not—support

SETC’s three-student model is a specific instructional design. It should not borrow stronger causal claims from research on “small groups” generally and pretend those studies tested this exact model.

The wider evidence is still useful because it helps identify plausible active ingredients and important cautions.

Small-group tuition

The Education Endowment Foundation (EEF) defines small-group tuition as one teacher, teaching assistant or tutor working with roughly two to five pupils. Its Teaching and Learning Toolkit reports positive average effects across the research it reviews, while also emphasising substantial variation between implementations.

EEF’s practical conclusions are especially relevant to SETC:

• support should be targeted to pupils’ specific needs;
• diagnostic assessment can help target that support;
• teaching quality may matter as much as, or more than, precise group size;
• staff preparation and high-quality interaction matter;
• subject matter and group composition can influence outcomes;
• small-group work should connect to the learner’s wider curriculum.

See: EEF — Small Group Tuition.

This evidence supports the proposition that well-targeted small-group tuition can be useful. It does not prove that a group of exactly three students is universally superior to one-to-one, four students or another well-designed format.

Collaborative learning

EEF’s collaborative-learning review distinguishes structured collaboration from simply placing learners in a group. It reports positive average effects but also lower evidence security and substantial variability.

The implementation guidance maps closely onto the design principles on this page:

• tasks need careful design;
• all learners need meaningful participation;
• teachers should model productive discussion;
• group composition matters;
• structured groups of about three to five learners can be promising;
• joint work should not allow one high-attaining learner to do the thinking for everyone.

See: EEF — Collaborative Learning Approaches.

SETC therefore uses peer contrast deliberately and preserves individual attempts. The group is not asked to “work together” vaguely.

Peer tutoring

EEF’s peer-tutoring evidence suggests that peer-supported learning can be useful, especially for reviewing or consolidating learning. That aligns with SETC’s decision not to ask students to become uncontrolled substitute teachers for brand-new material.

Peers can explain, compare, question and check against criteria after enough knowledge exists.

See: EEF — Peer Tutoring.

Feedback

EEF describes feedback as well evidenced overall, while emphasising that feedback quality and implementation matter. Useful feedback focuses on the task, subject and self-regulation and gives specific information about improvement. Feedback aimed at the learner as a person is generally less useful.

That supports SETC’s rule:

diagnose the decision, not the identity.

“This inference exceeds the evidence” is more actionable than “You are careless.”

See: EEF — Feedback and Teacher Feedback to Improve Pupil Learning.

Metacognition and self-regulation

EEF’s updated metacognition and self-regulation evidence emphasises explicit teaching of planning, monitoring and evaluation strategies, modelling of thinking, scaffolding and eventual learner self-regulation. It also notes that such strategies should be embedded in normal curriculum tasks rather than taught only as abstract “thinking skills”.

This is directly relevant to the three-student model’s final purpose: tutor questions should become learner questions.

See: EEF — Metacognition and Self-Regulation and the 2025 Metacognition and Self-Regulated Learning guidance.

Evidence caution

EEF itself cautions that Toolkit findings are summaries of average effects across studies and should support professional judgement rather than be treated as guarantees for one setting.

SETC should apply the same standard to its own model.

The page therefore makes a design argument:

three students can create a high-visibility environment for targeted English diagnosis, peer contrast, feedback and independence when the teaching is structured to use those conditions.

The strongest evidence that the model works for a particular learner remains the learner’s own progress across fresh, delayed and increasingly independent tasks.

What SETC should measure in its own implementation

A serious teaching model should become auditable over time.

Diagnostic precision

Can tutors identify repeatable weak links that predict later errors?

Repair success

Does the targeted intervention improve the trained decision on new examples?

Transfer

Does improvement survive unfamiliar material and reduced prompting?

Independence

Does tutor intervention decrease for already-trained decisions?

Group participation

Do all three learners retain meaningful cognitive work, or does one dominate?

Fit

How often does group composition need adjustment, and why?

Stage differences

Does the model operate differently in Primary, Secondary and JC contexts?

Subject-specific limits

Which English activities benefit most from peer contrast and which benefit less?

These are empirical questions for SETC’s own ongoing teaching practice. They should not be answered by branding.

Frequently asked questions

Why exactly three students?

SETC uses three because it aims to preserve close individual visibility while adding peer contrast. Three creates enough variation for comparison without requiring the tutor to monitor many simultaneous decision paths.

This is a design choice, not a scientific claim that three is universally optimal.

Is three better than one-to-one tuition?

Not in every situation. One-to-one can provide maximum individual attention and may be better for privacy, specialised objectives or large foundational gaps. Three-student tutorials add peer contrast, discussion and opportunities to observe alternative approaches.

The right choice depends on the learner and task.

Is three better than four or five students?

Not automatically. Well-designed groups of several learners can be effective. SETC’s three-student choice reflects its emphasis on continuous diagnostic visibility and individual feedback.

Will my child get one-third of the tutor’s attention?

Tutor attention should not be divided mechanically into equal minutes. At different moments, one student may need more direct intervention while the others perform meaningful independent or evaluative work.

The better question is whether each learner’s current English decisions remain visible and whether each receives the feedback and practice required.

What happens if the students are at different levels?

Some variation can be productive when the underlying mechanism is shared. The tutor can vary text complexity or extension. If the gaps are so large that the students require unrelated lessons, regrouping may be more appropriate.

What happens if one student is much faster?

Increase depth rather than worksheet volume. Ask the faster student to evaluate alternatives, generate counterexamples, justify decisions or transfer the principle to harder material.

What if one student dominates?

Use silent independent attempts, response order, rotating roles and tasks that require listening. Dominance is a group-design issue that the tutor should manage.

What if my child is shy?

Shyness alone does not make the format unsuitable. Written preparation and predictable turns can reduce speaking pressure. The tutor should distinguish quiet temperament from a learner being unable to participate safely or productively.

Do students copy one another?

They can if tasks are poorly designed. That is why independent commitment usually comes before comparison. Students write, select evidence or plan before discussion.

Does peer discussion teach mistakes?

It can. Peer explanations are not treated as automatically correct. The tutor monitors criteria and uses mistakes diagnostically.

Is the tutorial competitive?

Comparison should not become ranking. Students compare decisions against task criteria, not permanent ability labels.

How much homework should there be?

Enough to create useful retrieval and transfer evidence. More is not automatically better. Homework that is not reviewed or connected to diagnosis has low instructional value.

Can the model work online?

Yes, if independent evidence, participation and task visibility are preserved. Shared documents and recorded oral work can help. Technology also creates risks that need management.

Can students use AI?

Yes, when permitted and when the tool does not replace the target cognitive work. AI-supported practice should be followed by unaided transfer checks.

How quickly should marks improve?

There is no universal timeline. Improvement depends on starting state, skill, frequency, school task difficulty and how deep the repair is. Tutors should provide intermediate evidence rather than promising a fixed mark increase.

What if marks do not improve?

Review the diagnosis, task alignment, transfer, school requirements and group fit. If tuition activity increases but independent evidence does not improve, the teaching approach needs revision.

What if marks improve but the student still needs constant help?

That is an incomplete success. The model should test whether performance survives prompt removal.

What does “diagnosis” mean here?

Educational diagnosis in this page means identifying a repeatable learning or performance weak link from task evidence. It does not mean diagnosing medical or psychological conditions.

How is progress recorded?

Through small baseline tasks, error patterns, repair notes, fresh retests, delayed transfer and relevant school/examination evidence.

Should students be friends?

Friendship can help or distract. Group fit should be judged from actual learning behaviour.

Can siblings be in the same group?

Potentially, if age, stage, dynamics and learning needs fit. Family relationships can create additional interaction patterns that the tutor should consider.

Can Primary and Secondary students learn together?

Usually not for ordinary stage-specific tuition because language and curriculum demands differ substantially. Shared specialist workshops may occasionally make sense, but should be designed deliberately.

Can Secondary and JC students learn together?

Again, stage-specific paper work usually differs too much. Some advanced reading, rhetoric or vocabulary workshops can cross age boundaries when the mechanism and readiness align.

Does the tutor follow the school syllabus?

Stage and examination requirements matter, but SETC also teaches underlying mechanisms that transfer beyond one worksheet or school. The stage hubs and current official specifications guide paper-specific work.

What if the student is already doing well?

The group can shift from repair to extension: denser reading, more precise writing, rhetoric, research, Academic English or advanced reasoning.

What if the student has major gaps?

Begin at the first repairable prerequisite. If the gap makes the shared group task inaccessible, temporary one-to-one or another grouping may be more appropriate.

What makes this different from an ordinary small class?

The intended distinction is design: independent evidence before discussion, individual diagnosis, criterion-based peer contrast, targeted feedback, deliberate scaffold fading and transfer testing.

A three-person class without those features is simply small.

How do parents know the tutor is not overhelping?

Ask whether the learner can reproduce the skill on fresh work without tutor prompts. Improvement should increasingly survive outside the lesson.

When should tuition stop?

When the learner’s objective is stable enough that tutoring no longer adds sufficient value, or when another format better matches the next goal. A good teaching model includes an exit.

The teaching-model evidence hierarchy

SETC should interpret evidence in layers:

1. Wider research: what small-group, feedback, collaborative and metacognitive evidence suggests on average.
2. Local implementation evidence: what happens in SETC’s actual teaching model.
3. Individual learner evidence: what changes for this student across new tasks.

Do not jump from layer 1 to a guarantee at layer 3.

The model’s strongest defensible claim

Three students are enough to create meaningful peer contrast and few enough for a skilled tutor to maintain close observation. When lessons are designed around independent attempts, visible decisions, targeted feedback and transfer, that structure can support diagnostic English teaching without removing the learner’s responsibility for the work.

The model should be judged by whether it produces what it says it values:

visibility → diagnosis → repair → feedback → variation → transfer → independence.

Implementation quality: how the model survives beyond one excellent tutor

A teaching model is fragile if it works only when one unusually intuitive teacher is in the room. For SETC to use three-student tutorials as a durable institutional method, the important teaching decisions must be explicit enough to train, observe, discuss and improve.

This does not mean scripting every lesson. English teaching requires judgement. It means making the principles behind that judgement visible.

The tutor capability model

A tutor using the three-student system needs capability across several layers.

1. English subject knowledge

The tutor must understand the English mechanism being taught: reading, grammar, writing, oral, vocabulary, rhetoric, examination requirements or another domain.

No class-size design substitutes for subject expertise.

2. Diagnostic observation

The tutor must identify what evidence would reveal the learner’s process, notice divergence and avoid confusing a surface symptom with the first repairable cause.

3. Question design

The tutor needs questions that expose thinking without leading students step by step into the answer.

4. Feedback design

The tutor should be able to give information specific enough to change the next attempt and limited enough that the learner can act on it.

5. Group orchestration

The tutor manages participation, silence, role rotation, peer contrast and differentiation without letting the class become three simultaneous private lessons.

6. Scaffold control

The tutor knows when to model, prompt, wait, remove support and restore minimal support after a failed transfer.

7. Performance literacy

The tutor can interpret school and examination evidence without reducing learning to marks alone.

8. Metacognitive teaching

The tutor models planning, monitoring and evaluation and gradually transfers those functions to students.

9. Reference and evidence literacy

The tutor can verify uncertain language questions rather than teaching intuition as law. SETC’s Corpus, Dictionary & Reference Literacy owner provides the system.

10. Professional boundaries

The tutor recognises when a need is outside English tuition and does not invent clinical explanations.

Tutor training should use real student work

Generic workshops are not enough. Tutors can train by examining anonymised or simulated work and answering:

• What is the visible error?
• What are three plausible causes?
• What evidence would distinguish them?
• What is the smallest useful intervention?
• What fresh task would prove transfer?

This develops diagnostic judgement.

Lesson observation

Observing a three-student tutorial should focus on learning processes rather than tutor performance theatre.

An observer can record:

• percentage of lesson time students spend thinking/producing versus listening;
• whether all three make independent attempts;
• how often the tutor asks diagnostic questions;
• whether feedback differs by learner need;
• whether peer discussion has explicit criteria;
• whether support is faded;
• whether the lesson contains a retest.

The observer should not assume more tutor talk equals more teaching.

A tutor-observation rubric

Calibration across tutors

If several tutors use the model, they should periodically compare interpretations of the same student work.

Questions:

• Do we agree on the first weak link?
• Would we choose the same evidence?
• Which intervention would each tutor use?
• What would count as enough transfer?

Disagreement is useful because it exposes hidden assumptions in teaching judgement.

Curriculum alignment

Three-student tutorials should not become disconnected enrichment islands. The tutor needs awareness of:

• school stage;
• current curriculum;
• examination requirements where relevant;
• the learner’s school feedback;
• SETC’s deeper canonical English owners.

The tutorial can go deeper than school without becoming unrelated to school.

Canonical owner use inside tutoring

SETC’s website architecture can support teaching directly.

If the issue is inference, route to Reading & Comprehension.

If the issue is claim–evidence reasoning, route to Reasoning & Argument.

If the issue is register, route to World Englishes/Register owners.

If the issue is AI-assisted work, route to Digital & AI English.

The tutor should not recreate a private theory of English inside every lesson.

Lesson preparation

Preparation can be compact if the tutor knows the target.

Before class:

1. define the target decision;
2. choose one diagnostic task;
3. predict likely failure patterns;
4. prepare one narrow repair task;
5. prepare one transfer task;
6. identify extension for the fastest learner;
7. identify a load-reduction option for the learner who cannot access the task.

This is more useful than preparing thirty pages of worksheets.

Resource design

Materials for three students should support comparison.

Useful resources include:

• parallel passages;
• contrasting answer choices;
• multiple valid paragraph structures;
• worked examples with hidden decisions annotated;
• error sets grouped by mechanism;
• timed tasks with checkpoints;
• transfer tasks using unfamiliar surfaces.

Do not overproduce bespoke materials

Customisation can become inefficient if every student receives entirely different resources every week. Use a strong common task and individualise the observation, feedback or extension where possible.

Group continuity

Stable groups can develop efficient routines and trust. Constant regrouping resets social and instructional expectations.

Stability should not become rigidity. Regroup when evidence shows poor fit.

Joining an existing group

A new learner needs orientation:

• why independent attempts come first;
• why students sometimes compare wrong answers publicly;
• why the tutor may wait instead of helping immediately;
• how peer roles work;
• what respectful critique sounds like.

The culture of the group is part of implementation.

Leaving a group

Departure should be normal. A learner may:

• reach independence;
• move to a different stage;
• need another group level;
• need one-to-one temporarily;
• change learning objective.

Do not treat retention as the only sign of success.

Attendance

Because the model uses shared comparison, irregular attendance affects the group. However, a student should not be punished educationally for unavoidable absence.

Use concise catch-up focused on the mechanism, not full repetition of the missed class.

Replacement students

If one seat changes repeatedly, group dynamics and continuity suffer. SETC should consider compatibility before filling an empty seat merely to preserve commercial efficiency.

Scheduling

Three-student teaching can fail operationally if lessons are repeatedly shortened, rushed or interrupted. Enough time is needed for independent attempt, feedback and retry.

The model should not compress itself into “covering material” because the timetable is crowded.

Safeguarding and respectful interaction

Students should know that:

• errors may be discussed but not mocked;
• personal information remains appropriate to the learning context;
• peer feedback must address work or decisions;
• the tutor maintains professional boundaries;
• online sessions follow safe technology practices.

Psychological safety is not low challenge

A safe tutorial can still challenge weak reasoning directly.

“This evidence does not support your conclusion” is academically direct.

“You never understand inference” turns a task error into identity.

Challenge calibration

The task should be hard enough to reveal a decision but not so hard that every learner fails for unrelated reasons.

If vocabulary blocks comprehension entirely, the tutor cannot diagnose higher-level inference cleanly.

Desirable difficulty vs pointless difficulty

Variation, retrieval and reduced prompting can make learning appropriately difficult. Unclear instructions, irrelevant vocabulary or arbitrary time pressure can create noise.

The tutor should know which difficulty is being trained.

Implementation fidelity vs adaptation

SETC needs both.

Core principles should remain:

• independent evidence;
• diagnostic visibility;
• criterion-based peer contrast;
• targeted feedback;
• scaffold fading;
• transfer.

Surface routines can adapt by age, skill, medium and tutor style.

What should never become a rigid script

There is no requirement that every lesson use exactly three roles, one timing structure or one feedback format. Rigid routines can turn pedagogy into compliance.

Quality assurance through student work

The strongest audit evidence is not a tutor checklist. It is student work over time.

Sample anonymised sequences can show:

• baseline;
• feedback;
• revision;
• fresh task;
• delayed task.

This makes the model inspectable.

Quality assurance through learner explanation

Ask students what they are working on.

Weak answer:

“Comprehension.”

Stronger answer:

“I am learning to separate what the passage states from what I infer, because I was adding ideas outside the evidence.”

The student’s explanation reveals whether the lesson objective is cognitively available.

Quality assurance through parents

Parent feedback can reveal transfer:

• student starts homework with less prompting;
• student explains feedback more clearly;
• student reads more independently;
• student’s anxiety around tasks changes;
• school teacher comments align with tutorial evidence.

Parent observation is useful context, not a substitute for task evidence.

Quality assurance through school evidence

School scripts, rubrics and teacher comments can test whether tuition transfer reaches the actual curriculum environment.

Tutors should not assume their own internal exercises are the only valid evidence.

Model drift

As an organisation grows, a teaching model can drift.

Warning signs:

• “three students” remains in marketing while lessons become lecture-heavy;
• new tutors use worksheets instead of diagnosis because worksheets are easier to standardise;
• feedback becomes generic;
• transfer checks disappear under syllabus pressure;
• students remain dependent because parents value visible tutor help;
• group fit is ignored to keep seats full.

Preventing model drift

Use:

• tutor training on actual student decisions;
• lesson observation;
• student-work audits;
• shared owner pages;
• periodic calibration;
• parent communication standards;
• explicit release criteria.

Growth should not dilute visibility

SETC can scale its library, curriculum and tutor knowledge without increasing the three-student class size. The digital knowledge system can carry explanations; the tutorial uses human time for observation, feedback and transfer.

This division of labour is important.

A student can read the canonical explanation on SETC. The tutor does not need to recite it. Tutorial time can test what the student understands and can do.

Website and tutorial as one system

SETC’s long-term website architecture supports the teaching model:

• Atlas tells the tutor where the problem belongs;
• How English Works explains the mechanism;
• skill owners provide depth;
• stage hubs connect to curriculum;
• feedback and transfer owners explain repair;
• the tutorial provides observation and coached performance;
• independent tasks prove whether the learner can carry the skill away.

The website is not marketing surrounding the tuition. It becomes part of the learning infrastructure.

The three-student model in the eduKate federation

SETC owns the English teaching model. Local eduKate sites can apply the same principles in their own service contexts without recreating the English theory.

A Punggol or Sengkang family route can explain where tuition occurs or how local support is organised, then hand deep English mechanisms back to SETC.

This preserves federation ownership.

Implementation review cycle

At regular intervals, ask:

1. Are students doing enough independent cognitive work?
2. Are tutors identifying individual weak links?
3. Is peer contrast improving judgement?
4. Does feedback produce retries?
5. Do retries become transfer?
6. Is support fading?
7. Are some groups poorly composed?
8. Which resources create the clearest evidence?
9. Which lesson routines have become ritual rather than useful?
10. What should change next?

Researchable questions for SETC

As the teaching archive grows, SETC can investigate its own practice carefully without overstating causality.

Possible questions:

• Which reading errors are most stable across school stages?
• Which feedback types produce the strongest delayed transfer?
• How does peer contrast affect self-correction?
• At what point do checklists become unnecessary?
• Which group-composition variables predict poor fit?
• How often does one-to-one bridging improve later group participation?
• Which student-generated explanations best predict unseen performance?

Answers should use appropriate data and privacy safeguards.

Do not turn internal data into causal claims casually

If students improve after joining SETC, many factors can contribute: maturation, school teaching, additional practice, motivation, exam familiarity and tutoring.

Internal evidence can guide teaching and identify patterns. Strong causal claims require stronger designs.

The institutional standard for this model

A SETC three-student tutorial should be able to answer:

What can we see because the group is small?

What did we diagnose from that evidence?

What did we change?

What can the learner now do independently?

If the class cannot answer those questions over time, “three students” has become a number rather than a teaching model.

The final implementation principle

The smallest class is not automatically the best class.

The best use of a small class is to do educational work that would be difficult when student thinking is less visible.

For SETC, that work is:

observe → diagnose → compare → repair → retest → transfer → release.

How the three-student model changes as learners grow

The number of students can remain three while almost everything else changes: attention span, language knowledge, task length, independence, examination pressure, reasoning complexity and the tutor’s role.

A durable teaching model needs developmental progression. Otherwise “three-student tutorials” becomes one method imposed from Primary 1 to JC.

Early English and the beginning of formal learning

At the earliest stages, English grows through conversation, shared attention, stories, sound–print relationships, vocabulary, play and first writing. Diagnostic visibility is often oral and behavioural rather than examination-like.

What the tutor watches

• can the child follow and extend a spoken idea?
• can they hear and manipulate sounds?
• how do they approach unfamiliar print?
• do they use picture information as support or substitute for reading?
• can they retell a simple sequence?
• can they form and extend sentences?
• what happens when they do not know a word?

How peer contrast helps

Three children can hear different answers to the same story question, build vocabulary together and learn that another person may notice a different detail.

Competition should remain minimal. The aim is language participation and noticing.

Tutor role

The tutor models heavily, talks with rather than at learners, and keeps cycles short. Movement, drawing, manipulatives and shared reading can sit naturally inside the same diagnostic system.

The deeper route is Early English Learning.

Primary 1–2: build access and confidence

The model now begins to coordinate reading, grammar, vocabulary, first extended writing, listening and oral communication.

Priority

Build reliable access to instructions and text while keeping language meaningful.

Three-student task pattern

Short independent attempt → oral comparison → concrete explanation → immediate retry.

Useful peer contrast

• different decoding attempts;
• different sentence completions;
• different ways to retell a story;
• different vocabulary choices.

Risk

Turning the tutorial into premature examination drilling because the group is small enough to complete worksheets quickly.

Primary 3–4: transition into more formal English performance

Texts become denser, comprehension questions become more formal, composition demands increase and learners need stronger independent routines.

Priority

Move from “can do with support” toward “can identify what the task requires”.

Three-student task pattern

Students make independent evidence or planning decisions before discussion. Peer contrast becomes more explicit.

Useful roles

• evidence checker;
• scope checker;
• clarity checker.

Metacognitive language

Students begin naming their own strategies:

“I looked for the sentence that changes the character’s intention.”

not merely:

“I found the answer.”

Primary 5: increase integration

By Primary 5, isolated language skills need to work together more often. Reading affects writing. Vocabulary precision affects comprehension. Grammar control affects editing and composition.

Priority

Integrate without losing the underlying mechanisms.

Three-student use

Students can compare how one reading insight becomes three different writing choices, or how one vocabulary item behaves across reading, oral and composition.

Transfer

Begin stronger unseen-task testing. A method that works only on tuition worksheets is not enough.

Primary 6 and PSLE: reliability under pressure

The learning problem changes again. The learner needs both English capability and examination stability.

Priority

Preserve accuracy while coordinating papers, timing, stress and recovery.

Three-student use

Compare not only answers but examination decisions:

• question order;
• time allocation;
• what to check;
• when to move on;
• how to recover after a hard item.

Tutor role

Reduce unnecessary teaching of already-stable skills. Use time on bottlenecks and full-paper transfer.

The stage owner is PSLE English.

Common PSLE risk

The tutor may become more controlling as examination pressure rises—more hints, more model answers, more checking. This can produce short-term security and reduce independent execution.

The group should move in the opposite direction as the examination approaches: more realistic conditions, clearer recovery routines and less task-specific rescue.

Secondary 1: reset the learner for a new English environment

Secondary English introduces denser texts, broader writing forms and more independent decision-making.

Priority

Prevent Primary routines from becoming invisible constraints.

A student who succeeded through keyword matching may struggle with inference. A student who relied on fixed composition structures may struggle with new genres.

Three-student use

Compare how learners interpret new task types before assuming they understand them.

Tutor role

Teach fewer fixed templates and more transferable concepts: purpose, audience, evidence, paragraph job, tone, inference and self-checking.

Secondary 2: consolidate and expose emerging bottlenecks

This stage can appear stable while weaknesses accumulate beneath the surface.

Priority

Identify which lower-secondary habits will fail under upper-secondary complexity.

Three-student use

Use denser unseen material and more open writing choices. Observe whether students can select a method rather than follow one.

Secondary 3: complexity and pathway demands increase

Reading, summary, argument, situational writing and oral work become more demanding. Course pathways and examination expectations also matter.

Priority

Strengthen reasoning and transfer while maintaining language accuracy.

Three-student use

Peer challenge becomes more sophisticated:

• identify an unsupported claim;
• challenge a definition;
• compare two interpretations;
• test whether an example proves enough;
• adapt one message to a new audience.

Tutor role

Move increasingly from answer provider to discussion architect and diagnostic observer.

Secondary 4: final-year execution

The learner may know much more English than the examination result shows. Timing, question choice, recovery, editing and stress become more consequential.

Priority

Convert capability into reliable paper performance without narrowing English into paper tricks.

Three-student use

Run timed sections independently, then compare execution data:

• time checkpoints;
• error clusters;
• answer changes;
• unfinished sections;
• recovery after difficulty.

Release direction

By the final weeks, the tutor should increasingly behave like an external observer. Examination-day decisions belong to the student.

JC1 / pre-university: rebuild English as reasoning

General Paper and advanced English require students to read positions, evaluate evidence, use knowledge, build arguments and handle uncertainty.

Priority

Move beyond school-English routines into knowledge-mediated reasoning.

Three-student use

One student maps the author’s argument. One tests evidence. One builds a counterposition. Roles rotate before individual writing.

Tutor role

Challenge assumptions and definitions rather than supplying “good points”.

The stage owner is Advanced English / General Paper.

JC2: independence under severe constraints

Final-year GP work requires issue knowledge, question judgement, essay architecture, Paper 2 precision and reliable timing.

Priority

Integrate and release.

Three-student use

Students can review competing essay plans, defend question choice, synthesise source positions and audit one another’s evidence standards.

Tutor role

The tutor should no longer be the source of every example. Students need their own reading and knowledge systems.

University bridge: Academic English

Some learners continue into Academic English, where sources, citation, disciplinary conventions, research and seminars become central.

Priority

Transfer school-developed reasoning into independent research and communication.

Three-student use

Peer contrast can operate around:

• source interpretation;
• literature synthesis;
• research questions;
• academic paragraph architecture;
• presentation and Q&A.

The canonical owner is Academic English.

Extension beyond examinations

A learner who no longer needs examination repair can use the same tutorial format for:

• literature;
• rhetoric;
• creative writing;
• research;
• advanced vocabulary;
• public speaking;
• multilingual mediation;
• World Englishes;
• digital and AI literacy.

The model changes from remediation to intellectual apprenticeship.

How tutor talk should change over the years

How peer roles should change over the years

Young learners can use simple roles: listener, evidence finder, reteller.

Older Primary learners can check scope, evidence and clarity.

Secondary learners can challenge reasoning, register and structure.

JC learners can test assumptions, counterarguments and evidence quality.

The peer role becomes intellectually richer as language capability grows.

How feedback should change over the years

Early feedback can be concrete and immediate.

“You pointed to the right sentence. Now tell me which word shows he was worried.”

Later feedback can become more abstract:

“Your evaluation uses two criteria in different paragraphs. Choose one common criterion so the comparison becomes valid.”

The tutor’s language should grow with the learner.

How scaffolds should change over the years

Young learners may need visual organisers and sentence stems.

Older learners may need planning questions or criteria.

Advanced learners may need only a challenge:

“What would falsify your claim?”

The visible scaffold becomes increasingly internal.

How parent communication should change

For younger children, parents may need more information about routines, reading and home support.

For adolescents, communication should increasingly respect the learner’s ownership and avoid making the tutor-parent relationship a control system around the student.

By JC, the learner should usually be capable of explaining their own objectives and progress.

How measurement should change

Early learning evidence can include oral language, reading behaviour and first writing.

Primary evidence increasingly includes independent school-style tasks.

Secondary evidence includes genre control, unseen reading and timed performance.

JC evidence includes reasoning, synthesis, knowledge use and whole-paper execution.

The evidence must mature with the curriculum.

What remains constant from Primary to JC

Despite all these changes, six principles remain:

1. make thinking visible;
2. diagnose the first repairable cause;
3. use peer contrast without ranking identities;
4. give feedback that changes the next attempt;
5. reduce support as control improves;
6. prove learning on fresh, independent performance.

The longitudinal promise

A child should not experience ten years of tuition as ten years of being helped.

The progression should be:

help me see → help me choose → help me check → let me try → let me recover → let me own it.

That is the developmental logic of the three-student model.

Three-Student Tutorial Mastery Rubric

The model is mature when it can be evaluated without relying on slogans. The rubric below turns the teaching philosophy into observable conditions.

Learner maturity: four states

State 1: I need the tutor to show me

The learner lacks a usable model. Modelling and explanation are appropriate. The tutor should still keep the learner active through prediction and partial attempts.

State 2: I can do it when prompted

The learner knows the process but does not initiate it reliably. Prompts and checklists are useful. The next teaching objective is retrieval of the process itself.

State 3: I can do it independently in familiar conditions

The learner no longer needs the tutor for the basic decision, but performance may still depend on familiar task structure. Variation and integration become the priority.

State 4: I can adapt, recover and self-correct

The learner handles unfamiliar surfaces, recognises when the method needs modification, diagnoses common errors and seeks targeted help only where needed.

This is the release state for that mechanism.

Tutor maturity: four states

State 1: explain

The tutor knows the subject and can show correct methods.

State 2: diagnose

The tutor can distinguish why different students produce different errors.

State 3: orchestrate

The tutor designs shared tasks that preserve individual evidence, uses peer contrast and manages differentiated feedback without fragmenting the lesson.

State 4: release

The tutor knows when to reduce support, test transfer, stop teaching a stable target and recommend another format or no further tuition for that objective.

Parent-facing success criteria

A parent does not need to understand every pedagogical term. Over time, they should be able to observe some combination of:

• the child starts work with less prompting;
• the child can explain what went wrong and how to repair it;
• feedback becomes more specific and less repetitive;
• school work shows the trained skill without a tuition template;
• the child recovers from difficult questions more calmly;
• the tutor can explain what support is being removed, not only what is being added;
• the learner’s next goal changes because an older goal has become stable.

Programme-level success criteria

Across SETC, the model is healthy when:

• three-student classes remain genuinely three-student;
• tutors can articulate individual weak links;
• student work shows fresh-task transfer;
• peer comparison does not become ranking;
• groups are changed when fit is poor;
• parents receive mechanism-level progress information;
• tutors use canonical SETC owners instead of inventing conflicting mini-curricula;
• students increasingly manage their own English decisions.

The release statement

The strongest end-of-cycle statement is not:

“The student completed our programme.”

It is:

“The student can now perform this English decision on unfamiliar work with less support, can recognise common failure, and knows what to do when the first attempt breaks.”

That is what turns a three-student tutorial from a small class into a teaching system.

One-Lesson Audit: did this tutorial use the three-student model well?

After a lesson, the tutor can review the session in two minutes. The audit is deliberately short because an implementation tool that is too expensive will not survive ordinary teaching.

1. Independent evidence

Did every student make at least one meaningful decision before hearing the others?

If no, the tutor has weak evidence about individual capability.

2. Visible thinking

Could I explain how each student arrived at one important answer, plan or performance decision?

If no, the next lesson needs a task that exposes process rather than only outcome.

3. Individual diagnosis

Can I name the current weak link for each learner in task language rather than personality language?

Examples:

evidence scope, paragraph purpose, retrieval speed, pronoun reference, checking behaviour.

Not:

careless, lazy, weak, anxious writer.

4. Useful peer contrast

Did the students learn something from seeing another approach, or were they simply waiting for the correct answer?

Peer interaction should alter judgement, not merely distribute answers.

5. Targeted feedback

Did feedback tell each learner what decision to change next?

If the feedback could have been copied onto all three pages unchanged, ask whether the diagnosis was sufficiently individual.

6. Retry

Did the learner get another chance to perform the corrected decision?

Explanation without retry leaves the repair untested.

7. Productive struggle

Did I allow enough thinking time before rescuing?

If I supplied answers rapidly, I may have protected lesson pace at the cost of retrieval and independence.

8. Participation

Did all three students do substantive cognitive work?

Speaking time can differ. Thinking responsibility should not disappear for one learner.

9. Scaffold direction

Did support increase because the task exposed a genuine gap, remain stable because it is still needed, or decrease because control improved?

If scaffolds never change, the teaching may not be responding to evidence.

10. Transfer plan

What fresh task will test whether today’s learning survives?

Name it before the lesson is forgotten.

11. Group fit

Did the common task remain useful for all three?

If one learner spent most of the session waiting or working on unrelated material, review differentiation or grouping.

12. Tutor necessity

Which action did I perform today that the learner should eventually perform themselves?

That answer becomes a future fading target.

The audit score is not the point

SETC should not turn these questions into a superficial 12-point performance score. Their function is diagnostic. One weak dimension can matter more than ten strong ones.

If students produced excellent answers but no independent evidence existed, visibility is the next repair. If visibility was strong but the tutor gave no retry, feedback implementation is the next repair. If all teaching was excellent but the group no longer shares a useful task, grouping is the next repair.

The same philosophy used with students applies to the teaching model itself:

find the first repairable cause, change it, and gather fresh evidence.