IGCSE Composition Feedback | Turn a Marked Script into a Better Unseen Response

IGCSE composition feedback matters only if it changes what a learner can do on a different prompt. A comment that produces a prettier redraft but disappears on the next unseen task has improved one script. It has not yet improved examination performance.

This guide is built around that distinction. It takes the marked composition as the starting point, not the finishing point. The reader job is to move through a complete chain: inspect the evidence, interpret the comment, compare possible causes, choose one bounded repair, test that repair in revision, remove the original script, and then ask whether the same decision appears independently on a fresh narrative or descriptive task.

Maren, Iona and Leonie return as SETC’s resident trio. Maren is especially vulnerable to feedback that sounds artistic but is too vague to execute: “develop this more”, “make the atmosphere stronger”, “show rather than tell”. Iona insists that every comment be connected to evidence and an explanation of what the reader can actually infer. Leonie asks the examination question: can the improved decision survive without the teacher, the model, the original paragraph or extra time?

The article uses original SETC teaching scripts, comments, revisions and transfer tasks. They are not Cambridge candidate scripts, examiner reports, official candidate responses or Cambridge-awarded marks. Where feedback depends on the examination criteria, use the correct official materials for the year in which the candidate will sit Cambridge IGCSE First Language English 0500.

The 50-second router

If you have just received a marked composition and do not know what to do next, start with The feedback translation rule. It turns broad comments into observable writing decisions.

If you keep making the same mistake after correcting it, go to Why correction is not transfer and the first-weak-link sections.

If your teacher says “develop”, “be more descriptive”, “vary sentences”, “improve vocabulary” or “check accuracy”, use the comment-decoder sections. Each shows how to move from the label to evidence, cause and a bounded repair.

If you can produce a much better redraft but still perform inconsistently in tests, go to the transfer-probe sections. They distinguish assisted revision from independent retrieval.

If you are a teacher or tutor, use the case files and feedback-design sections to check whether your comments identify a mechanism that can survive a new prompt.

The examination-year rule

Cambridge currently separates the 2024–2026 and 2027–2029 syllabus cycles for First Language English 0500. For examinations from 2027, Paper 2 still includes Composition, while the assessment-objective wording and marking criteria are updated. The official 2027 specimen mark scheme continues to divide Composition into Content and Structure and Style and Accuracy, with 16 and 24 marks respectively.

This page therefore separates two things:

durable feedback mechanics — evidence, diagnosis, bounded repair, retrieval, transfer and independence;

year-specific assessment language — the exact descriptors, task instructions and official wording used in the relevant syllabus cycle.

Do not create a hybrid mark scheme by combining phrases from 2026 and 2027. Do not assume a teacher comment has the same assessment meaning across qualifications merely because both use words such as “structure”, “style” or “accuracy”. Feedback has to remain attached to the correct task and criteria.

Official starting point: Cambridge IGCSE First Language English 0500. For examinations from 2027, use the official specimen Paper 2 and specimen Paper 2 mark scheme alongside the syllabus.

The feedback translation rule

Every useful composition comment should eventually answer five questions.

1. Where is the evidence? Which sentence, paragraph, pattern or whole-script feature triggered the comment?
2. What is the mechanism? What writing decision is producing the effect?
3. Why does it matter? Which part of the composition performance is limited?
4. What is the smallest useful repair? What can the learner change without rewriting everything?
5. How will transfer be tested? What fresh task will show that the learner can reproduce the improved decision independently?

Consider the teacher comment: Develop your ideas more.

That comment may be correct and still be unusable.

Possible mechanisms include:

• narrative events arrive without consequences;
• a descriptive paragraph lists details without changing the picture;
• the central object appears but never influences action;
• a character reacts emotionally but does not make a decision;
• the writer repeats the same idea using different adjectives;
• paragraphs are individually strong but do not build on each other;
• the ending introduces a new meaning too late to be earned.

Those are different problems. They require different repairs.

A translated comment might become:

Evidence: paragraphs three and four both show Nina feeling worried after the lost card. Neither paragraph changes her available options.

Mechanism: emotional repetition without causal development.

Repair: replace one reaction paragraph with a decision that creates a new consequence.

Transfer: on a fresh narrative prompt, mark the point where each middle paragraph changes what the protagonist can do next.

Now “develop more” has become executable.

Feedback is not correction

Correction tells the learner what is wrong and often supplies or implies a repaired form.

Feedback is broader. It can help the learner notice a quality gap, understand its cause, choose a better decision and test whether that decision becomes available later.

If a teacher changes:

“The crowd was very loud and chaotic and everyone was moving everywhere.”

to:

“Voices overlapped beneath the roof while queues bent around one another and trolley wheels struck the metal barriers.”

the sentence is improved, but the learner may have learned nothing beyond the teacher’s superior phrasing.

A feedback sequence asks the learner to identify what changed:

• abstract labels became observable evidence;
• “loud” became overlapping voices;
• “chaotic” became conflicting physical movement;
• the scene now contains spatial relationships;
• the vocabulary is more precise because the observation is more precise.

Then the learner must produce a different crowded scene without the model sentence.

Correction is useful. It simply does not prove transfer.

Why correction is not transfer

A student can understand a correction immediately and still repeat the original error two days later.

That happens because recognition is easier than retrieval.

When the teacher points at a comma splice and says, “These are two independent clauses,” the learner may understand perfectly. On a fresh timed composition, no teacher points at the sentence. The learner has to notice the boundary while also planning content, selecting vocabulary, maintaining viewpoint and managing time.

The repair therefore needs at least three states:

supported repair: the problem is identified for the learner;

independent repair: the learner finds and corrects the problem without a cue;

independent production: the learner produces a fresh sentence or paragraph without recreating the same error.

The third state is the examination target.

The four-feedback distinction

Not all feedback should be treated as the same kind of message.

1. Outcome feedback

Examples: 29/40; Level 5 Content and Structure; Style and Accuracy lower than last time.

This tells you where the performance landed. It does not necessarily tell you why.

2. Evidence feedback

Examples: “The final third contains five sentence-boundary errors”; “paragraphs four and five repeat the same descriptive job”; “the prompt object disappears after the opening”.

This tells you what happened on the page.

3. Mechanism feedback

Examples: “When you increase syntactic complexity, clause boundaries become unstable”; “your narrative middle expands through extra events rather than through consequences”; “you label atmosphere when close observation becomes difficult”.

This proposes a cause that can be tested.

4. Transfer feedback

Examples: “On the next unseen narrative, make each middle paragraph change an available choice”; “in the next description, ban emotional atmosphere labels and create mood only through observable relationships”; “under time, protect the final eight minutes for completion and accuracy”.

This turns diagnosis into a future performance condition.

Strong feedback systems move from outcome toward transfer.

The first weak link

A marked script can contain twenty visible problems. Training them all at once usually produces noise.

Find the earliest mechanism that creates later damage.

Example:

The ending is abrupt.

The climax is compressed.

The middle contains three unnecessary events.

Those events exist because the original plan tried to cover a week of story time.

The first weak link is scope selection at planning, not “write better endings”.

Another example:

The last paragraph has spelling slips, vague vocabulary and repetitive sentences.

All three appear only because the learner has four minutes left.

The first weak link may be execution time, not spelling, vocabulary or sentence variety.

Feedback becomes efficient when it moves upstream.

The feedback ledger

Keep one page per marked composition with these fields:

Prompt:

Conditions: timed / untimed / unseen / model-supported / prompted / revised.

Outcome: mark or level profile if available.

Evidence: what the script actually shows.

Possible cause A:

Possible cause B:

Evidence that separates A from B:

First weak link:

Bounded repair:

Fresh transfer task:

Transfer result: secure / partial / not yet.

This ledger prevents the same comment from being rediscovered as if it were new every month.

Decode the comment before obeying it

Short feedback comments are compressed messages. A teacher writes “develop”, “awkward”, “too general”, “vocabulary”, “structure” or “check punctuation” because there is not enough room to explain the full diagnosis beside every sentence.

The student’s job is not to obey the label blindly. It is to expand the comment back into a testable mechanism.

“Develop this”

Possible meanings:

• the event happens but its consequence is missing;
• the image appears once and never changes;
• the paragraph states an emotion without showing what causes or changes it;
• the idea is relevant but occupies too little space relative to its importance;
• the writer jumps from premise to ending without an intermediate decision;
• the description names a place but does not establish relationships among details.

Before revising, ask which one is visible in the script.

“Be more descriptive”

This does not automatically mean “add more adjectives”. It may mean the reader cannot visualise spatial relationships, cannot hear or see the relevant action, or cannot tell which details matter.

Possible repairs include:

• replace atmosphere labels with concrete evidence;
• move from wide view to a close detail;
• show how light, sound or movement changes across the scene;
• give the observer a stable physical position;
• select one object whose condition reveals use, neglect or transition;
• remove irrelevant sensory details that dilute the dominant effect.

“Show, don’t tell”

This famous comment is often overused. Some information should be told directly because examination composition has limited space.

The useful question is: which moment deserves expansion?

If a character is nervous about knocking on a door, behaviour may carry the feeling better than the sentence “she was nervous”. If the story needs to tell us that three weeks passed, direct summary is efficient.

Translate the comment into: “This emotional or perceptual moment carries enough weight to deserve observable evidence rather than a label.”

“Improve vocabulary”

Possible meanings:

• too many vague verbs such as went, got, did, looked;
• repetition of one adjective;
• wrong collocation;
• imprecise noun choice;
• forced rare words;
• topic vocabulary is too thin;
• the student knows words but retrieves them too slowly under time.

Those problems should not receive the same worksheet.

If the issue is forced rarity, the repair may be simpler language. If the issue is vague verbs, the repair may be observation plus verb selection. If the issue appears only under time, the repair may be retrieval practice rather than new vocabulary study.

“Vary your sentences”

Do not respond by alternating long and short sentences mechanically.

Possible underlying problems:

• every sentence begins with the subject;
• every idea is expressed as a separate simple sentence;
• coordination is overused;
• subordination is absent;
• fragments are overused for drama;
• long sentences are not controlled;
• sentence shape does not respond to pace or emphasis.

A useful repair identifies a grammatical relationship first. If the learner needs to show cause, condition, concession or time, practise that relationship on new content.

“Structure”

This comment can refer to at least six different levels:

• whole-piece order;
• paragraph sequence;
• paragraph job duplication;
• proportion;
• transition;
• ending relationship to the opening.

Ask the teacher, tutor or yourself: “Which two paragraphs would you point to as the clearest evidence of the structural problem?”

“Check accuracy”

Do not begin with a full grammar textbook.

Sort the errors:

• sentence boundaries;
• tense consistency;
• subject–verb agreement;
• pronoun reference;
• dialogue punctuation;
• apostrophes;
• spelling;
• capitalisation;
• word form;
• run-on syntax under ambition.

Then ask which pattern carries the greatest cost or appears most often.

The rival-diagnosis rule

The first explanation is not always the right explanation.

If a student writes a rushed ending, at least four causes are possible:

1. they cannot write endings;
2. the plot became too large;
3. the opening took too long;
4. the student lost time because they stopped repeatedly to search for vocabulary.

Each cause produces the same visible symptom: weak ending.

Feedback becomes more accurate when you compare rival causes.

How to compare causes

Use additional evidence from the same script and from other conditions.

If untimed endings are also weak, the problem may be structural knowledge.

If untimed endings are strong but timed endings collapse, execution becomes more likely.

If the ending is weak only when the plot contains many characters, scope becomes more likely.

If the student spends five minutes replacing ordinary words with rare ones, retrieval behaviour may be part of the cause.

Diagnosis should therefore ask, “What else would I expect to see if this cause were true?”

The minimum-repair rule

A repair should be large enough to teach the mechanism and small enough that the learner can see cause and effect.

Too small:

Teacher changes one comma and the student copies it.

Too large:

Student rewrites the entire composition after receiving twelve different comments.

Better:

Student repairs four sentence-boundary errors of the same type, explains the rule, then writes four fresh sentences using the same relationship.

Better:

Student rewrites one descriptive paragraph so each sentence contributes to a chosen transformation, then writes a new paragraph in a different setting.

Better:

Student replaces one disconnected narrative event with a consequence of the previous choice, then plans a different story using the same causal principle.

Bounded repair makes learning visible.

The feedback burden

More comments do not automatically create more learning.

A heavily marked page can overwhelm the learner’s attention. Ten circles, five marginal notes, three arrows, a score and a paragraph of end comments may all be individually accurate while collectively making the next action unclear.

Use feedback hierarchy.

Priority 1: first weak link that limits the whole response.

Priority 2: recurring error pattern with high cost.

Priority 3: one strength to preserve during repair.

Everything else can be recorded but does not have to become immediate practice.

For example:

Priority 1: narrative middle does not develop causally.

Priority 2: comma splices in complex sentences.

Preserve: precise dialogue and concise opening.

That is enough for a strong next training session.

Feedback should protect strengths as well as repair weaknesses

A common failure occurs when the learner fixes a weakness by destroying an existing strength.

A student with concise, focused writing is told to “develop more” and responds by adding unnecessary backstory.

A student with accurate simple sentences is told to “vary sentences” and begins writing unstable long constructions.

A student with a coherent small-scale narrative is told to “make it more exciting” and adds unrelated dramatic events.

Every repair should therefore include a preservation clause:

Change: make the middle causal.

Preserve: keep the current two-character scope.

Change: increase syntactic range.

Preserve: keep sentence boundaries accurate.

Change: strengthen descriptive atmosphere.

Preserve: keep the current spatial route clear.

Improvement is not rebuilding from zero. It is controlled change.

Case File 1 — Narrative development: “Nothing really changes in the middle”

Original SETC practice prompt

Write a story in which someone is given responsibility for something they did not want.

Student first draft

Leonie stared at the dog lead in her father’s hand.

“Just for the afternoon,” he said. “I have to go back to the office.”

The dog, Pepper, sat beside the front door with his tongue out, looking completely pleased with the arrangement.

Leonie was not.

She had planned to meet Iona at the library. They had a history presentation due on Monday and had agreed to finish the slides. Leonie looked at her phone.

“Can’t you ask Mrs Tan?”

“She’s away.”

“What about Adam?”

“Football.”

Her father put the lead in her hand.

“Two walks. Food at five. Don’t let him into the study.”

Then he left.

Leonie was annoyed. Pepper followed her everywhere. When she went to the kitchen, he went to the kitchen. When she went upstairs, he sat at the bottom and whined. Leonie tried to work on her laptop, but Pepper kept making noises. She gave him a biscuit and told him to be quiet.

After a while Pepper barked at the window. Leonie told him to stop. Then he barked again. She looked outside and saw another dog going past. Pepper ran around the living room. Leonie was getting more annoyed.

At three o’clock she took him outside. Pepper pulled on the lead. Leonie nearly dropped her phone. A cyclist went past and Pepper barked. A child wanted to stroke him. Leonie said yes. Pepper was excited.

When they got home Leonie was tired. She tried to do her presentation again, but Pepper scratched at the study door. Leonie remembered that her father had said not to let him in. She moved him away.

At five she gave Pepper his food. Pepper ate it quickly. Leonie sent Iona a message saying she was sorry about the library.

At six her father came home.

“How was he?” he asked.

“Annoying,” Leonie said.

Her father laughed. “But you managed.”

Leonie looked at Pepper. She realised that responsibility was difficult but sometimes you have to do things you do not want to do.

The teacher’s original feedback

“You need more development in the middle. The story becomes repetitive. Try to make something happen that changes Leonie’s responsibility.”

This feedback is useful because it already points beyond “add detail”. The next job is to test what “repetitive” means.

Evidence before diagnosis

Iona maps the middle paragraphs:

• Pepper follows Leonie; she is annoyed.
• Pepper barks; she is annoyed.
• Pepper pulls outside; she is annoyed.
• Pepper scratches at the study door; she moves him away.
• Pepper eats; she misses the library.

The surface events change, but Leonie’s available choices and understanding barely change. Her emotional state repeats.

The final insight—responsibility is difficult but necessary—is stated at the end rather than produced by a meaningful choice.

Rival diagnosis A: “Not enough exciting events”

Maren’s first response is to add a dramatic event: Pepper runs into the road.

That would certainly create change. But it may also push the student toward a familiar rescue-story pattern and larger physical stakes than the prompt needs.

The problem is not lack of excitement. It is lack of consequence.

Rival diagnosis B: “Character needs more emotion”

The draft already names annoyance several times. Adding frustration, panic and relief might increase emotional vocabulary without improving the plot.

The mechanism remains unchanged unless emotion leads to a decision.

Stronger diagnosis: responsibility is present but never tested

The prompt gives a useful central concept: unwanted responsibility.

The story should therefore create a moment when Leonie has an easier selfish option and a harder responsible option.

That test can remain small.

Pepper scratches at the forbidden study door. Instead of simply moving him away, Leonie hears a faint alarm from inside. She has been told not to enter. She also knows her father stores work equipment there. Now the rules conflict:

obey the exact instruction versus protect what she has been left responsible for.

That is development because the meaning of responsibility becomes more complex.

Bounded repair: rewrite only the study-door sequence

Original:

“She tried to do her presentation again, but Pepper scratched at the study door. Leonie remembered that her father had said not to let him in. She moved him away.”

Revised:

Leonie opened her laptop again. Pepper crossed the landing and stopped outside the study.

He scratched once.

“No.”

He scratched again, harder.

From behind the door came a thin electronic chirp.

Leonie waited. The sound repeated.

Her father’s instruction returned exactly: Don’t let him into the study.

That was not the same as Don’t go into the study.

She clipped the lead onto Pepper’s collar and tied it to the stair rail. Then she opened the door.

A desk lamp had fallen against a stack of papers. Its bulb pressed into the top sheet, leaving a brown circle in the paper. The chirp came from the smoke detector above the door.

Leonie pulled the plug from the wall and carried the papers into the corridor.

Pepper barked once, triumphantly.

“You are still annoying,” she told him.

But when she returned to her laptop, she moved his bed beside her chair.

What changed?

The repair did not make the story longer everywhere. It created one decision with consequences.

Responsibility now changes meaning:

• initially: an unwanted chore;
• then: a set of rules;
• then: judgement when rules conflict;
• finally: a relationship Leonie chooses to manage more actively.

The revised dog-bed detail is stronger than the original moral statement because changed behaviour demonstrates the change.

Preservation clause

Keep the small domestic scale.

Keep only a few characters.

Keep the concise opening.

Do not add fire engines, injury or a dramatic rescue. The mechanism is choice under responsibility, not spectacle.

Revision is not yet transfer

Leonie can produce the improved paragraph because the teacher and SETC have identified the exact weak point.

That proves understanding under support.

Now remove the dog, the house, the study, the parent and the forbidden room.

Fresh transfer probe

Prompt: Write a story in which someone agrees to look after an object that belongs to another person.

Constraints:

• no animals;
• no parents;
• no lost phones;
• no theft;
• no fire or accident;
• the central object must create a conflict between two reasonable responsibilities.

Possible independent plan

Maren agrees to look after a borrowed musical instrument during a school event.

Instruction 1: keep it in the locked rehearsal room.

New problem: another student needs the same room for an emergency sound check, and the lock is malfunctioning.

Choice: protect the object by refusing access, or protect the event by finding a supervised alternative.

Consequence: Maren must create a third solution under pressure rather than simply follow one instruction.

Transfer criterion

The learner succeeds if the new story contains a moment where responsibility changes from simple obedience into judgement, without reproducing the dog-story wording or events.

If the student writes another story about guarding a forbidden room, surface dependence remains.

Delayed transfer

Three or four days later, use a different prompt:

“Write a story in which a small promise becomes difficult to keep.”

Do not mention responsibility. Do not remind the learner of the previous repair.

If the learner independently creates a point where two obligations conflict and a choice changes later consequences, the feedback has begun to transfer.

Case File 1 lesson

The comment “develop more” became useful only after the system identified the repeated mechanism.

The final chain is:

comment → evidence → rival causes → causal diagnosis → one bounded repair → fresh prompt → delayed probe.

That chain is the model for the rest of this article.

Case File 2 — Descriptive attention: “You keep telling me the atmosphere instead of making me see it”

Original SETC practice prompt

Describe a market just before heavy rain begins.

Student first draft

The market was crowded and very noisy. Many people were walking around everywhere. The atmosphere was hot and uncomfortable, and everyone looked impatient because the weather was getting worse.

The sky was dark and threatening. Huge black clouds covered everything. The stalls were colourful and busy. Sellers shouted loudly while customers rushed around. There were many delicious smells from the food stalls, and the whole place felt chaotic.

It was extremely humid. People were sweating. A woman fanned herself with a newspaper. A man carried several bags. Children looked tired and irritated. The market was becoming tense because everyone knew a storm was coming.

Suddenly the wind became stronger. Plastic bags flew through the air. The sellers started becoming worried. Some pulled covers over their goods. Others shouted at their helpers. Everyone moved faster and the place became even more chaotic.

Then one large drop of rain landed on the ground. Another followed. Everyone began running for shelter as the rain poured down.

The teacher’s comment

“Too general. You tell me that the market is noisy, crowded, tense and chaotic, but many details could fit any busy place. Choose a clearer viewpoint and let specific changes create the atmosphere.”

Maren hears only the first two words: too general.

Her first instinct is to add more adjectives.

Iona stops her.

“The draft already has plenty of adjectives. The problem is that the adjectives are doing jobs the observations should do.”

Evidence map

Underline atmosphere labels:

• crowded;
• very noisy;
• hot;
• uncomfortable;
• impatient;
• dark and threatening;
• busy;
• chaotic;
• extremely humid;
• tired and irritated;
• tense;
• worried;
• even more chaotic.

Now underline details that actually let the reader infer those conditions:

• a woman fans herself with a newspaper;
• plastic bags fly;
• sellers pull covers over goods;
• one large drop lands.

The ratio tells us something. Most of the paragraph’s energy goes into naming the effect rather than constructing it.

Rival diagnosis A: “Needs more senses”

A standard worksheet might tell the student to add sound, smell, sight, touch and taste.

But the draft already mentions sound, smell, heat, clouds and movement. The issue is not missing sensory categories.

The issue is selection and relationship.

Five disconnected senses can still produce a list.

Rival diagnosis B: “Needs stronger vocabulary”

The student could replace “dark” with “brooding”, “noisy” with “cacophonous” and “humid” with “oppressive”. The paragraph might sound more ambitious while remaining equally general.

The feedback target is not lexical rarity. It is observational precision.

Stronger diagnosis: the observer has no route

The draft moves among sky, stalls, people and weather without a stable path.

The student therefore reaches for global labels because there is no chosen sequence of attention.

Give the observer one location: beside a fruit stall near the covered edge of the market.

Choose one transformation: ordinary bargaining → collective preparation for rain.

Now details can be selected according to that transformation.

Bounded repair: rewrite only the first three paragraphs

Revised version:

From the fruit stall at the edge of the market, the roof looked lower than it had ten minutes earlier.

Beyond its metal lip, the sky had lost its blue without becoming fully grey. Light lay flat across the oranges and mangoes, leaving no shine on their skins. The seller picked up the handwritten price cards one by one and slid them beneath the weighing scale although nobody had asked her to.

The aisle was still full, but movement had changed. Customers no longer stopped in the centre to compare bags. They stood closer to the stalls, leaving a narrow channel between them. A man with two grocery bags checked the open road, then changed both bags to one hand so he could unfold an umbrella.

At the cooked-food stall opposite, steam no longer rose straight above the pots. It leaned toward the entrance.

The fruit seller noticed it too.

She looked once at the sky and pulled a roll of clear plastic from beneath the counter.

Somewhere behind the roof, metal struck metal: one shutter, then another.

The first wind entered the aisle carrying the smell of wet concrete from the road before any rain reached the market itself.

What changed?

The revision does not tell the reader that the atmosphere is tense.

Instead, ordinary behaviour changes:

• price cards are removed early;
• customers alter where they stand;
• a man prepares an umbrella;
• steam changes direction;
• plastic is unrolled;
• shutters begin closing;
• the smell of wet concrete arrives before the rain.

The storm exists through effects.

The hidden structural repair

The student thinks the improvement is “better detail”.

Iona points out the deeper repair: the details are now ordered.

Viewpoint begins at the fruit stall.

Focus moves outward to sky.

Then down the aisle.

Then across to the cooked-food stall.

Then returns to the fruit seller.

Then sound and smell extend beyond the visible roof.

The route makes the description coherent.

Preservation clause

Keep the approaching-rain premise.

Keep the market crowded.

Keep the original useful idea of sellers covering goods.

Do not add a dramatic accident, argument or escape. The composition is developing through collective preparation, not plot.

Fresh transfer probe

Prompt: Describe a sports ground in the minutes before play is stopped because of weather.

Constraints:

• do not use a market;
• do not use food;
• do not use umbrellas;
• do not call the atmosphere tense, chaotic, threatening or dramatic;
• choose one fixed observer position;
• show at least three ordinary behaviours changing before the weather fully arrives.

Possible transfer decisions

The groundskeeper checks the boundary rope twice.

Players look toward the same corner of sky between deliveries.

Spectators begin moving coats from seats to laps.

A scoreboard operator places a cover within reach without fitting it yet.

Wind turns one flag before the others.

The principle transfers: atmosphere is made from relationships and changes, not labels.

Delayed transfer probe

Several days later, use:

“Describe a place preparing for an important arrival.”

Do not mention atmosphere or viewpoint in the instruction.

Success is stronger if the learner independently chooses a stable observational position and lets changed behaviour reveal anticipation.

Why the redraft alone can mislead

After reading the revised market paragraph, a student may simply copy its surface architecture: object close-up, man doing something, sound in distance, smell at end.

That is not the learning target.

The transferable principle is:

choose a dominant change → place the observer → select behaviours and details that measure the change → move attention coherently → allow atmosphere to emerge.

Case File 2 feedback card

Original comment: too general; show the atmosphere.

Evidence: repeated abstract labels; few prompt-specific relationships.

Rival causes rejected: missing senses; weak advanced vocabulary.

Mechanism: no stable viewpoint or hierarchy of attention.

Repair: fixed observer + one transformation + selected behavioural evidence.

Preserve: market premise and approaching weather.

Transfer: different setting, no atmosphere labels, changed behaviours must carry the effect.

When “be more descriptive” should mean “delete”

Sometimes stronger description requires fewer words.

Consider:

“The gigantic enormous black clouds were extremely dark and terrifying, filling the whole frightening sky with a sinister and ominous presence.”

Deletion can improve this faster than addition.

A more precise version might be:

“The cloud line erased the pale strip of sky above the roofs.”

The second sentence gives a relation the reader can see.

Feedback should therefore distinguish density from precision. More modifiers can reduce clarity if they all compete to name the same effect.

The attention hierarchy

Strong descriptive feedback often needs to answer: what should the reader notice first, second and last?

A practical hierarchy is:

1. orientation;
2. dominant change or contrast;
3. one or two load-bearing details;
4. secondary detail that extends the scene;
5. final concentration or transformed view.

This is not a compulsory template. It is a diagnostic tool for descriptions that feel like piles of observations.

If every detail has equal weight, the reader has to invent the hierarchy. Feedback can teach the writer to choose it.

Case File 3 — Language control: “Your ideas are strong, but your accuracy falls when sentences become ambitious”

Original SETC practice prompt

Write a story in which someone receives information they wish they had known earlier.

Extract from the student’s timed first draft

Maren had waited outside the office for nearly twenty minutes, the corridor was empty except for a cleaner at the far end who was pushing a trolley slowly towards the lifts.

When Mrs Vale finally opened the door, she didn’t say come in, instead she looked at the envelope in Maren’s hand and asked where she had got it from.

“It was in my locker” Maren said, “I thought you left it there.”

Mrs Vale’s expression changed, although she smiled it was obvious that something was wrong and Maren suddenly remembered the phone call from yesterday which she hadn’t answered because she had been on the bus and her battery was almost dead.

Inside the office there were two chairs, a small table covered in folders, and the window which was open even though it had started raining, making the papers on the table move.

Mrs Vale told her that the letter was supposed to have been collected three days earlier, Maren had been chosen for the exchange programme but because they thought she had declined the place it had been offered to someone else.

Maren felt shocked, furious and devastated, she wanted to ask why nobody had told her, why the message had not been sent again, and why the letter had been put in a locker she almost never used.

She looked down at the envelope which was now useless and wished she had known earlier.

The teacher’s feedback

“Strong situation and clear consequence. Your main limit is sentence control: several long sentences join complete clauses with commas, dialogue punctuation is unstable, and tense/reference become less secure when you pack too much into one sentence. Repair the grammar without flattening the ideas.”

Why “write shorter sentences” is an incomplete repair

Leonie can fix every visible error by splitting aggressively:

Maren waited outside the office. It was twenty minutes. The corridor was empty. A cleaner pushed a trolley. The cleaner went towards the lifts.

The result is technically safer but loses relationships that the original writer was trying to express: simultaneity, contrast, cause, concession and relative importance.

The feedback target is not length. It is relationship control.

Evidence map

Look at the first sentence:

“Maren had waited outside the office for nearly twenty minutes, the corridor was empty except for a cleaner at the far end who was pushing a trolley slowly towards the lifts.”

Two independent clauses are joined by a comma.

The writer wants to do two things:

• establish duration;
• establish emptiness and slow movement.

Possible repairs:

“Maren had waited outside the office for nearly twenty minutes. The corridor was empty except for a cleaner at the far end, pushing a trolley slowly towards the lifts.”

or:

“As Maren waited outside the office, the corridor remained empty except for a cleaner pushing a trolley slowly towards the lifts.”

The choice depends on emphasis. Grammar correction is also a meaning decision.

Rival diagnosis A: “The student does not know punctuation”

Perhaps. But the same student may punctuate simple sentences accurately all the way through.

The errors cluster specifically when:

• two complete thoughts are combined;
• dialogue and action share a sentence;
• subordinate clauses stack;
• the writer adds an explanatory clause late in a long sentence.

That pattern suggests partial knowledge plus overload, not total ignorance.

Rival diagnosis B: “The student needs less ambition”

Permanent simplification would protect accuracy but may cap stylistic range.

The stronger goal is controlled ambition.

Sentence surgery 1 — comma splice

Original:

“When Mrs Vale finally opened the door, she didn’t say come in, instead she looked at the envelope in Maren’s hand and asked where she had got it from.”

Repair:

“When Mrs Vale finally opened the door, she did not say, ‘Come in.’ Instead, she looked at the envelope in Maren’s hand and asked where she had got it.”

What changed?

• direct speech is punctuated as speech;
• “instead” begins a new independent sentence rather than being attached by a comma;
• the final “from” is optional and removed for cleaner phrasing.

Sentence surgery 2 — dialogue punctuation

Original:

“It was in my locker” Maren said, “I thought you left it there.”

Repair:

“It was in my locker,” Maren said. “I thought you left it there.”

The student should not merely copy the punctuation. They should explain the pattern: the reporting clause follows the first spoken sentence; the second spoken sentence is separate.

Sentence surgery 3 — overloaded concession

Original:

“Mrs Vale’s expression changed, although she smiled it was obvious that something was wrong and Maren suddenly remembered the phone call from yesterday which she hadn’t answered because she had been on the bus and her battery was almost dead.”

The sentence tries to carry:

• change of expression;
• contrast between smile and concern;
• Maren’s inference;
• memory of a call;
• reason she missed it;
• battery condition.

That is too many relationships for the student’s current control.

Repair:

“Mrs Vale’s expression changed. Although she smiled, something in her face made Maren think of yesterday’s missed call. She had ignored it on the bus because her battery was almost dead.”

The revision preserves complexity of meaning while distributing it across manageable sentence structures.

Sentence surgery 4 — run-on cause and consequence

Original:

“Mrs Vale told her that the letter was supposed to have been collected three days earlier, Maren had been chosen for the exchange programme but because they thought she had declined the place it had been offered to someone else.”

Repair:

“The letter should have been collected three days earlier. Maren had been chosen for the exchange programme, but when the school received no reply, the place was offered to someone else.”

This is not merely more correct. It is clearer about causality.

Bounded repair exercise

The learner does not rewrite the full composition yet.

They take six sentences from the original and classify the intended relationship:

• time;
• cause;
• contrast;
• addition;
• condition;
• reported speech/dialogue.

Then they repair each sentence in at least two grammatical ways.

For example, cause can be expressed by:

• because;
• so;
• a separate sentence plus a causal adverb;
• a participial phrase, when genuinely controlled;
• reordering cause before consequence.

The student learns choices, not one template.

Preservation clause

Keep the strong central situation.

Keep the clear consequence of lost opportunity.

Keep the contrast between ordinary office details and the importance of the news.

Do not simplify the plot merely to simplify the grammar.

Fresh transfer probe 1 — sentence level

Without looking at the original, write six new sentences about a delayed train. Each must express a different relationship:

• time;
• cause;
• contrast;
• condition;
• speech plus action;
• two related independent ideas without a comma splice.

Fresh transfer probe 2 — paragraph level

Prompt: A student discovers that the venue for an important event has changed.

Write 180–220 words. Use at least:

• one controlled subordinate clause;
• one correctly punctuated line of dialogue;
• one sentence that deliberately remains short;
• one sentence combining two related ideas without a run-on.

Do not mention exchange programmes, lockers, missed calls or envelopes.

Fresh transfer probe 3 — composition level

Several days later:

“Write a story in which someone learns an important fact later than they should have.”

No sentence checklist is provided.

Now the learner must retrieve the control independently while also managing plot and time.

How to judge transfer

Surface success: the six repaired original sentences are correct.

Near transfer: the delayed-train sentences are correct when the relationship is named.

Stronger transfer: the paragraph uses varied relationships accurately with only broad constraints.

Examination transfer: the fresh complete composition shows improved sentence control without a cue.

Feedback should move through those states.

Case File 3 feedback card

Original comment: strong ideas, unstable sentence control.

Evidence: comma splices, dialogue punctuation errors, overload in long sentences.

Rival causes: total punctuation ignorance versus overload under complexity.

Diagnosis: grammatical relationships are understood conceptually but not executed reliably under syntactic load.

Repair: classify relationship → repair multiple ways → generate new sentences → use in new paragraph.

Preserve: plot quality and conceptual ambition.

Transfer: fresh complete story without sentence prompts.

Accuracy feedback should name patterns, not shame error

“Careless” is rarely a useful diagnosis.

Some errors are slips. Others are predictable products of load, incomplete automatisation or a rule that has not been understood.

Replace:

“You are careless with punctuation.”

with:

“Your sentence boundaries are secure in simple sentences but become unstable when you join independent clauses. We will practise that relationship under increasing time pressure.”

The second comment gives a learner something to test.

The complexity ladder

For students whose accuracy collapses as sentences become ambitious, increase difficulty gradually.

1. two short independent sentences;
2. coordination with and/but/so where appropriate;
3. subordination for one clear relationship;
4. sentence with one embedded phrase;
5. sentence combining subordination and modification;
6. paragraph where structures are selected freely;
7. timed paragraph;
8. timed full composition.

Do not jump from correction straight to a full mock and conclude that the repair failed.

Training needs an intermediate bridge.

Case File 4 — Completion under pressure: “Your ending is weak” is sometimes a timing diagnosis

Original SETC practice prompt

Write a story in which a person has to make a choice before it is too late.

Conditions

Timed, unseen, independent practice.

The student planned for nine minutes, drafted for most of the remaining time and reached the final paragraph with fewer than four minutes left.

Student first draft — compressed reconstruction

Iona receives a call from a former teammate asking her to attend a selection trial that begins in ninety minutes.

She has already promised to help her younger cousin prepare for a school presentation.

The opening establishes both commitments clearly.

The next section shows Iona debating whether the trial could restart a part of her life she had abandoned after an injury. She finds her old sports bag, notices the faded number on it, and remembers the last match she played.

Then she travels to her cousin’s house.

Her cousin has not prepared the presentation. They argue. Iona helps organise the slides. The cousin’s printer stops working. They walk to a stationery shop. The shop is crowded. Iona sees a former teammate there. They talk about the selection trial. Iona checks the time.

On the way back, a bus is delayed. The cousin becomes upset. Iona reassures him. They finally print the presentation. Iona receives another call saying the trial registration will close in twenty minutes.

The final paragraph reads:

“Iona suddenly realised that she still wanted to go to the trial and that she had already helped her cousin enough. She told him that she had to leave and ran to the sports centre as fast as she could. When she got there the doors were closing but she managed to enter just in time. She learned that sometimes you have to choose your own future.”

The teacher’s feedback

“The ending is rushed and too convenient. The final choice should carry the story’s pressure, but here it is summarised in a few sentences. Look at how much space the middle events consume.”

The obvious repair

Maren wants to rewrite the final paragraph beautifully.

She adds:

• the sound of Iona’s shoes on the pavement;
• a description of the sports centre;
• the final seconds before registration closes;
• a reflective sentence about opportunity.

The ending improves.

The diagnosis remains incomplete.

Evidence before diagnosis

Leonie marks the narrative by functional time and word space.

Opening problem: efficient.

Backstory: useful but slightly long.

Cousin conflict: relevant.

Printer failure: secondary obstacle.

Walk to shop: transition.

Crowded shop: setting expansion.

Former teammate conversation: repeats the trial reminder.

Bus delay: another obstacle.

Cousin distress: repeats responsibility pressure.

Second call: final deadline reminder.

Choice and consequence: compressed into the ending.

The story has allocated the largest amount of space to obstacles before the central choice.

Rival diagnosis A: “Student cannot write endings”

Check untimed work.

In untimed practice, the same student writes controlled endings that return to earlier details and show changed behaviour.

That weakens the diagnosis.

Rival diagnosis B: “The student panics in the last minutes”

Possibly, but panic is not the whole cause.

The student arrived at the last minutes with too much story still unfinished.

Pressure was created upstream.

Stronger diagnosis: scope inflation

The original plan contains too many intermediate obstacles.

The printer problem, shop trip, teammate encounter and bus delay all compete for words. Several serve the same function: make the trial harder to reach.

A stronger structure could preserve only one external obstacle and one relational obstacle.

For example:

• relational obstacle: cousin genuinely needs help;
• external obstacle: the printer failure forces a decision about how much more help Iona can reasonably give.

The shop, teammate and bus can disappear.

Bounded repair: restructure before rewriting the ending

New paragraph jobs:

1. Call creates opportunity and deadline.
2. Existing promise to cousin creates conflict.
3. Iona helps cousin and discovers the task is larger than expected.
4. Printer fails; time threshold arrives.
5. Iona must choose whether to stay until every problem is solved or hand the cousin a workable plan and leave.
6. Choice creates emotional consequence with cousin.
7. Ending shows whether Iona reaches the trial, but the real climax is the boundary she sets.

Now the centre of gravity moves toward the choice.

Revised decisive sequence

The printer clicked twice and stopped.

Iona checked the screen. Paper feed error.

Her cousin stared at her.

“We can go to the shop.”

Iona looked at the time.

Thirty-one minutes.

“If we go now, I miss registration.”

“But I need the handouts.”

She almost reached for her coat anyway. That was what helping had meant for the last hour: solve the next problem, then the next one.

Instead she opened a blank message on her phone.

“Email the slides to your teacher tonight. Present from the laptop tomorrow. The handouts can wait.”

Her cousin’s face tightened.

“So you’re leaving?”

Iona picked up the old sports bag.

“I’m leaving you with a presentation you can give.”

He looked at the laptop, then at her.

“That’s not the same as helping.”

“No,” Iona said. “It’s not the same as doing everything.”

She left before she could make the sentence softer.

What changed?

The revision makes the choice itself the climax.

Whether Iona reaches the trial remains important, but the deeper change is her definition of responsibility.

The ending can now be smaller:

She reaches the centre after registration has technically closed. A staff member checks the list, sees that she called earlier and allows her to wait for the organiser. The result of the trial does not need to be shown.

Or she arrives too late and still recognises that making the choice mattered. Either ending can work if prepared honestly.

Preservation clause

Keep the competing responsibilities.

Keep the old sports bag as a physical reminder of the opportunity.

Keep the cousin relationship.

Remove repeated obstacles that perform the same structural job.

The timing ledger

Leonie adds one more layer to feedback: when did the student reach each stage?

Record approximate checkpoints during practice:

• question selected;
• plan complete;
• opening complete;
• central turn reached;
• climax reached;
• draft complete;
• final check begins.

If several scripts show the climax arriving too late, the issue is not an isolated ending.

Fresh transfer probe

Prompt: Write a story in which someone must leave before a task is completely finished.

Constraints:

• maximum three significant characters;
• one central external obstacle only;
• the decisive choice must occur before the final quarter of the planned story;
• the ending may not introduce a new problem;
• reserve a fixed final checking interval during the timed attempt.

Near-transfer drill

Before writing a full response, plan three versions of the story in six bullet points each.

Delete any event that does not:

• increase pressure;
• change a choice;
• create a consequence;
• prepare the ending.

Delayed transfer probe

A week later:

“Write a story in which an opportunity disappears while someone is deciding what to do.”

No event-count limit is provided.

Success means the learner independently controls scope and reaches the decisive turn with enough writing time remaining.

Case File 4 feedback card

Original comment: ending rushed and convenient.

Visible symptom: climax and consequence compressed into final lines.

Rival causes: weak ending knowledge; panic; general writing weakness.

Evidence: untimed endings stronger; timed middle contains repeated obstacles.

Diagnosis: scope inflation consumes the time and word budget required for the central choice.

Repair: remove duplicate obstacles; move decision earlier; protect completion time.

Preserve: central conflict and character relationship.

Transfer: fresh story with independently controlled event count and earlier turn.

Feedback on endings should often begin with the opening

An ending problem may be caused by the promises the opening makes.

If the first 120 words introduce:

• four characters;
• two timelines;
• a mystery;
• a family conflict;
• a competition;
• an unexplained object;

the ending inherits all of those obligations.

Before teaching a more elegant final paragraph, ask whether the story has promised more than 350–450 words can responsibly deliver under the relevant examination conditions.

The completion hierarchy

When time is running short in practice, teach students to protect these in order:

1. complete the central causal or descriptive movement;
2. reach an ending that belongs to the task;
3. preserve sentence boundaries and basic accuracy;
4. check high-risk errors;
5. polish only if time remains.

A complete controlled piece usually has greater examination value than an ornate first two-thirds followed by collapse.

Do not teach a magic minute allocation as official fact

Training can use time budgets, but they are coaching choices unless the examination board specifies otherwise.

One student may need slightly longer to plan and write faster. Another may plan quickly but require a larger checking window.

Feedback should therefore use observed execution data:

“In your last three practices, you reached the central turn with less than ten minutes remaining.”

That is stronger than:

“Everyone must finish paragraph four at minute X.”

The goal is a reliable execution system, not ritual compliance.

The feedback cycle: from marked script to future performance

A useful feedback system has a beginning and an exit. Without an exit, the learner can spend days inside one old composition.

Use this seven-stage cycle.

Stage 1 — Preserve

Keep the original script exactly as submitted. Do not silently correct it before analysis.

The original is evidence of the student’s independent state under those conditions.

Stage 2 — Interpret

Translate comments into evidence and mechanisms.

If the teacher writes “weak structure”, identify where the structure weakens. If the teacher writes “awkward”, identify the grammatical or lexical relationship that creates awkwardness.

Stage 3 — Prioritise

Choose one first weak link and, at most, one secondary recurring pattern.

The purpose is not to ignore other weaknesses. It is to create a sequence the learner can execute.

Stage 4 — Repair

Revise the minimum amount needed to understand the mechanism.

Sometimes that is four sentences. Sometimes two paragraphs. Sometimes a new six-point plan.

Stage 5 — Explain

The learner states what changed and why.

If they cannot explain the repair without pointing to the teacher’s wording, understanding may still be shallow.

Stage 6 — Transfer

Use a fresh surface with the same underlying decision.

Change setting, relationship, object, viewpoint, prompt form or genre conditions enough that copying cannot solve the task.

Stage 7 — Delay

Retest after time has passed and the original script is no longer active in working memory.

The strongest evidence is independent use under realistic examination conditions.

The feedback exit rule

Stop revising the old script when all four conditions are met:

1. the learner can name the first weak link;
2. the learner can demonstrate the repair in the original context;
3. the learner can explain the decision without teacher wording;
4. a fresh prompt is ready for transfer testing.

Do not wait until the old composition is perfect.

Perfection in one familiar script can create false confidence.

The one-comment lesson

Some of the best feedback lessons begin with only one comment.

Example comment:

“Your paragraph stays in the same emotional state.”

A 45-minute lesson can then do:

1. locate the paragraph;
2. identify the repeated state;
3. compare two possible causal revisions;
4. rewrite six lines;
5. explain the mechanism;
6. plan a different story where a decision changes the emotional state.

The learner leaves with one usable principle rather than twelve half-processed corrections.

Feedback granularity: whole script, paragraph, sentence, word

Comments fail when they operate at the wrong level.

Whole-script feedback

Use when the problem is:

• task fit;
• scope;
• genre control;
• overall structure;
• proportion;
• ending relationship;
• viewpoint consistency across the piece.

Paragraph-level feedback

Use when the problem is:

• paragraph job duplication;
• weak development;
• transition;
• local descriptive route;
• missing consequence;
• overextended backstory.

Sentence-level feedback

Use when the problem is:

• boundary control;
• subordination;
• coordination;
• reference;
• tense;
• dialogue punctuation;
• rhythm or emphasis.

Word-level feedback

Use when the problem is:

• precision;
• collocation;
• register;
• spelling;
• word form;
• unnecessary repetition;
• forced rarity.

Do not use a word-level repair for a whole-script problem.

Replacing ten adjectives will not fix a plot that has no causal development.

The feedback funnel

Move from large scale to small scale unless there is a severe language problem blocking meaning.

A useful order is:

1. task fit;
2. whole-piece development;
3. paragraph sequence and proportion;
4. sentence control;
5. lexical precision;
6. surface proofreading patterns.

Why?

Because polishing a sentence that will later be deleted is inefficient.

Because vocabulary cannot rescue the wrong task.

Because punctuation practice is valuable but does not tell a student where the story is going.

This order is not absolute. If sentence control is so weak that meaning breaks down, language may need immediate priority. The funnel is a default, not a law.

The feedback queue

Not every comment has to be trained immediately.

Keep three queues.

Now: the first weak link for the current cycle.

Next: an important recurring pattern that should be trained after the first repair stabilises.

Watch: a weakness seen once but not yet proven to be a pattern.

Example:

Now: narrative scope.

Next: comma splices under complexity.

Watch: two spelling errors on unfamiliar words.

This protects the learner from treating every red mark as equal.

The evidence threshold before calling something a pattern

One error may be a slip.

Two similar errors may suggest a pattern.

Repeated errors across different prompts and conditions provide stronger evidence.

Do not diagnose “weak tense control” from one accidental shift if three other narratives remain stable.

Do not diagnose “poor vocabulary” because one topic produced repeated nouns.

Build enough evidence to justify the intervention.

The teacher-question ladder

Feedback should gradually remove the teacher’s thinking.

Level 1 — direct diagnosis

“Paragraphs three and four repeat the same emotional state. Replace one with a decision and consequence.”

Level 2 — narrowed question

“Which two paragraphs perform the same job?”

Level 3 — criterion question

“Where does the narrative stop developing?”

Level 4 — self-diagnosis

“What is the first weak link in this script?”

Level 5 — independent monitoring

No teacher question during the task. The learner identifies the pattern during checking or after completion.

The goal is not to keep feedback forever. It is to internalise the questions feedback has been asking.

Feedback latency: immediate versus delayed

Immediate feedback is useful when:

• the learner is practising a new grammatical pattern;
• an error will be repeated many times if left unchecked;
• the task is explicitly guided practice;
• misunderstanding is blocking progress.

Delayed feedback is useful when:

• you need to observe independent performance;
• the learner must practise self-monitoring;
• the task simulates examination conditions;
• you want to reduce dependence on tutor interruption.

During an exam simulation, constant correction destroys the measurement.

During initial sentence-control teaching, withholding every correction may waste practice.

Feedback timing should match the training state.

Feedback before writing can become hidden prompting

If the teacher says, “Remember to use a repeated object with changed meaning,” immediately before an unseen composition, and the student does it, the performance is supported.

That does not make the writing worthless. It changes what the result proves.

Record the support.

Then fade it:

1. specific reminder;
2. general reminder to plan development;
3. no reminder;
4. delayed self-assessment after the task.

Feedback becomes part of learning only when the learner eventually initiates the decision.

The redraft gap

Track the difference between first draft and revision.

A large redraft improvement can mean:

• the learner understands feedback well;
• the learner has strong potential with time;
• the learner benefits from teacher diagnosis;
• the learner’s main weakness is self-monitoring;
• the learner’s main weakness is time pressure.

Do not conclude automatically that the original skill is now mastered.

The redraft gap is useful data.

The transfer gap

Now compare revised performance with a fresh independent task.

Example:

Original narrative: middle development weak.

Revised narrative: strong after teacher feedback.

Fresh narrative: middle development weak again.

The learner has revision competence but incomplete transfer.

Training should now reduce cues and increase retrieval rather than give another detailed explanation.

The condition gap

Compare the same skill under different conditions.

Untimed: sentence control secure.

Timed paragraph: mostly secure.

Timed full composition: accuracy drops in final third.

This suggests the grammar knowledge exists but is vulnerable to total-task load.

The next intervention should approximate that load more closely.

Feedback should end with a prediction

Before the next task, write a prediction:

“If our diagnosis is correct, then on a fresh prompt with only two significant characters, the narrative should reach its decisive turn earlier and the ending should no longer be compressed.”

or:

“If the sentence problem is caused by syntactic load rather than basic punctuation knowledge, accuracy should remain strong when one complex relationship is required but decline when several are stacked.”

Predictions make diagnoses testable.

If the prediction fails, revisit the diagnosis.

The feedback success hierarchy

Use five states.

State 1 — corrected: the original error is repaired with help.

State 2 — explained: the learner can explain the repair.

State 3 — reproduced: the learner uses the principle on a closely related new task.

State 4 — transferred: the learner selects it on a substantially different task.

State 5 — stabilised: the learner maintains it across multiple unseen timed tasks.

Do not call State 1 mastery.

Teachers and tutors: design feedback that can survive the next prompt

A composition comment should ideally produce a future action that still makes sense when the setting, characters and topic change.

Compare two comments.

Surface comment: “Make the market description more vivid.”

Transferable comment: “Your paragraph labels the crowd as chaotic but does not show how movement is organised. On the next description, choose one observer position and show three behaviours changing as the scene changes.”

The second comment can transfer to a station, sports ground, theatre or school hall.

The three-part end comment

For a complete IGCSE composition, an efficient end comment can contain three parts.

Preserve: one strength that must survive revision.

Repair: one first weak link.

Transfer: one condition for the next fresh task.

Example:

Preserve: Your central character’s motive is clear and the opening reaches the problem quickly.

Repair: In the middle, each new event arrives from coincidence. Replace one event with a consequence of the character’s previous choice.

Transfer: On the next unseen narrative, underline the sentence where each middle paragraph changes the protagonist’s options.

This comment is shorter than a page of corrections and gives a clear training path.

Separate assessment from coaching

Assessment asks: what does this submitted script demonstrate?

Coaching asks: what should the learner do next?

Do not inflate the assessment mark because you know the learner could fix the problem with guidance.

Do not make coaching punitive because the mark is low.

A script may receive a modest mark and still reveal an easy-to-fix first weak link.

Another script may receive a strong mark but reveal unstable transfer that requires careful work.

The comment-to-task rule

Every major feedback comment should be convertible into a practice task.

Comment: “Your ending is rushed.”

Weak task: write another full story.

Stronger task: plan three six-event narratives and delete any event that performs the same function as another; then write only the decisive sequence and ending.

Comment: “Vocabulary is repetitive.”

Weak task: learn twenty sophisticated words.

Stronger task: take one setting and build lexical sets for movement, texture and sound; write two new paragraphs using precise ordinary and advanced words naturally.

Comment: “Description is too general.”

Stronger task: describe a new place without atmosphere adjectives; all mood must be inferable from behaviour, spatial relation, light, sound or material condition.

If a comment cannot generate a task, it may still be too vague.

The evidence-first conference

A five-minute teacher–student feedback conference can follow this order.

1. Student summarises the composition’s intended effect.
2. Teacher asks student to identify one strength with evidence.
3. Teacher asks student to identify one limit with evidence.
4. Teacher adds or corrects the diagnosis.
5. Together they choose the first weak link.
6. Student states the bounded repair.
7. Teacher gives the fresh transfer task.

The student should speak before the teacher gives the final diagnosis whenever the learning stage allows it.

This reveals whether the learner can read their own writing.

Do not turn feedback into prose replacement

Teachers with strong writing ability can accidentally become ghostwriters.

The student writes:

“The hall was empty and quiet.”

The teacher writes a polished replacement.

The student copies it.

The final redraft looks better. The learner’s decision-making may be unchanged.

Prefer prompts that preserve ownership:

“What evidence would prove the hall has just emptied?”

“Which sound would become noticeable only after people leave?”

“Choose one object that shows recent use.”

Then let the student write the sentence.

When direct modelling is appropriate

There are times when a teacher should model.

If the learner has never seen how to repair a comma splice, demonstrate.

If the learner cannot distinguish an atmosphere label from observable evidence, show a contrast pair.

If the learner’s paragraph has no coherent route, think aloud through one example.

Then fade the model quickly:

1. teacher model;
2. joint attempt;
3. student attempt with question prompts;
4. student independent attempt;
5. fresh transfer task.

The model is a bridge, not a permanent dependency.

Feedback frequency

Marking every full composition in maximal detail can be inefficient for both teacher and learner.

Use varied feedback modes.

Full-profile feedback: complete script, both assessment dimensions, first weak link, transfer task.

Single-focus feedback: one skill only, such as causal development or sentence boundaries.

Self-assessment check: student marks first, teacher responds to diagnosis.

Peer evidence check: partner identifies evidence without assigning a mark.

Silent performance sample: no intervention during writing; analyse later.

Different modes answer different questions.

Feedback and marking workload

A teacher does not need to annotate every error to teach accurately.

For a recurring grammar pattern, mark the first two or three examples and ask the learner to find the rest.

For paragraph-job duplication, label two repeated jobs and ask the learner to map the remaining paragraphs.

For vague vocabulary, circle a small representative sample and ask the learner to classify why each word is vague.

This shifts some cognitive work back to the student.

Parents: turn the marked script into a calm next step

Parents often see the total mark first.

The more useful sequence is:

1. What did the teacher say to preserve?
2. What is the first weak link?
3. What small repair has been assigned?
4. When will the child test it on a fresh prompt?

Avoid turning feedback into a second marking session at home.

If the teacher has identified causal development, the parent does not need to add vocabulary, handwriting, spelling and “more exciting ideas” on the same evening.

Parent questions that help

Ask:

• “Can you show me the sentence or paragraph the comment refers to?”
• “What do you think caused it?”
• “What are you changing first?”
• “What are you keeping because it already works?”
• “How will you know this is fixed on a different topic?”

These questions reinforce evidence and transfer.

Parent questions that often hurt

Avoid:

• “How many marks did you lose?”
• “Why didn’t you remember what your tutor taught you?”
• “Can you just memorise a better story?”
• “Why can’t you write like the model answer?”
• “What vocabulary will get you more marks?”

Those questions pull attention toward score recovery, blame or imitation rather than mechanism.

When parent help becomes hidden support

A parent may brainstorm the plot, suggest the climax, supply vocabulary and correct sentences before the child submits practice.

The resulting composition can be excellent.

It is not an independent performance sample.

Record support honestly. Then create a later task without it.

Support is not bad. Invisible support makes diagnosis bad.

Tutors: distinguish lesson success from exam readiness

A student may leave a tutorial able to repair every marked paragraph.

That shows lesson learning.

Exam readiness requires the student to:

• notice the problem without the tutor;
• select the right strategy;
• execute it on new content;
• do so under realistic time and task load.

Tutors should therefore schedule uncued performance samples.

The tutor-prompt ledger

Record prompts during practice.

Examples:

• “What does this paragraph need to change?”
• “Have you reached your central turn?”
• “Check the subject of this sentence.”
• “Where is the observer standing?”

If the same prompt is needed repeatedly, that is evidence of cue dependency.

Fade it deliberately.

Feedback should become self-talk

The mature learner eventually carries compact internal questions.

Maren:

“What does this paragraph change?”

Iona:

“What evidence makes the reader infer that?”

Leonie:

“Can I still finish this scope with control?”

Those questions are more valuable than memorising the wording of old teacher comments.

Feedback and motivation

Accurate feedback does not require exaggerated praise.

It does require preserving agency.

Compare:

“You always rush your endings.”

with:

“In the last three timed narratives, your central turn arrived in the final ten minutes. We will reduce middle-event count and test whether that moves the turn earlier.”

The second statement is specific, temporary and testable.

It describes a performance pattern, not a permanent identity.

Feedback and high-performing students

Strong students need feedback too, but the job changes.

Instead of repairing basic coherence, feedback may target:

• overwriting;
• forced sophistication;
• predictable prepared structures;
• subtle viewpoint leakage;
• uneven risk-taking;
• small differences between a strong and a fully controlled ending;
• stability across unfamiliar prompts.

Do not respond to a high mark by saying only “keep it up”.

Ask what still prevents the next level or greater reliability from becoming the better description.

Feedback and lower-performing students

Do not give the entire mark scheme at once.

Choose the first mechanism that unlocks the greatest amount of writing.

For one student, that may be sentence boundaries.

For another, it may be keeping one narrative problem instead of five.

For another, it may be writing complete paragraphs around one job.

The same examination criteria can lead to different teaching sequences.

The feedback contract

A productive feedback culture can use a simple contract.

The teacher promises:

• comments will point to evidence;
• major comments will be prioritised;
• strengths worth preserving will be named;
• assessment and coaching will be distinguished;
• feedback will eventually fade.

The learner promises:

• not to copy corrections without explanation;
• to attempt the bounded repair;
• to keep the original script;
• to complete the fresh transfer task;
• to report honestly when the repair does not transfer.

Feedback becomes a shared investigation rather than a red-pen verdict.

Measure transfer by distance from the original feedback

A repair becomes more convincing as the learner moves farther from the conditions that produced it.

Transfer distance has at least four dimensions.

1. Surface distance

How different is the new content?

Low distance: dog responsibility story → cat responsibility story.

Higher distance: dog responsibility story → borrowed musical instrument.

Higher still: responsibility story → promise story where two obligations conflict but responsibility is never named.

The greater the surface change, the less likely success comes from copying.

2. Cue distance

How much of the feedback is still visible?

Low distance: teacher says, “Make the middle causal.”

Higher distance: learner sees a card saying “development”.

Higher still: no cue; learner diagnoses the need during planning or checking.

3. Time distance

How long has passed since the repair?

An immediate retry tests short-term availability.

A retry several days later tests retrieval after forgetting has begun.

A later unseen timed task tests whether the decision has become part of the learner’s working system.

4. Load distance

How much else must the learner manage at the same time?

Low load: repair one sentence.

Higher load: write a paragraph.

Higher still: complete a full composition while choosing a prompt, planning, drafting, monitoring time and checking accuracy.

True examination transfer requires success at high load.

The four-distance transfer matrix

Record each transfer attempt across the four dimensions:

Surface: similar / changed / substantially changed.

Cue: direct / partial / absent.

Delay: immediate / short delay / longer delay.

Load: sentence / paragraph / full timed composition.

A repair may look secure in one cell and fragile in another.

Example:

Sentence boundaries:

• similar surface, direct cue, immediate, sentence level — secure;
• changed surface, partial cue, next day, paragraph level — secure;
• changed surface, no cue, one week, timed composition — unstable in final third.

The correct conclusion is not “fixed” or “not fixed”. The control is developing but not yet stable under total-task load.

False progress signal 1 — the beautiful redraft

A revised composition can be dramatically better because:

• the teacher supplied the diagnosis;
• the learner has unlimited time;
• a model answer is nearby;
• the same sentences have been reread many times;
• the learner knows exactly where the problem is;
• the surface content no longer has to be generated.

That improvement is real. It is evidence of revision competence.

It is not sufficient evidence of unseen-task competence.

False progress signal 2 — the near-identical retry

A learner repairs “atmosphere through changed behaviour” in a market and then succeeds on a “shopping centre before rain” task.

The transfer is promising but close.

Change more variables before declaring the mechanism stable.

False progress signal 3 — the teacher’s question is still doing the thinking

The tutor asks, “What does this paragraph change?”

The student immediately improves the paragraph.

This proves the learner can respond to the cue.

It does not yet prove the learner will generate that question independently.

Fade the cue and retest.

False progress signal 4 — higher marks from easier prompts

A mark rises after feedback, but the new prompt happens to match a familiar setting or prepared plot.

Do not attribute the entire improvement to the feedback.

Use several varied unseen tasks before inferring a stable gain.

False progress signal 5 — accuracy improves because ambition disappears

After sentence-boundary feedback, a student writes only short simple sentences.

The error rate falls.

Control may not have expanded. The learner may simply be avoiding the grammatical relationship that caused difficulty.

Retest with appropriately scaffolded syntactic range.

False progress signal 6 — model phrasing replaces the old weakness

The learner stops writing vague descriptions but begins recycling polished phrases from exemplars.

Surface quality rises while independence remains weak.

Ban the model’s wording and change the setting.

What “fixed” should mean

A feedback target is reasonably stable when the learner can:

1. recognise the mechanism in their own writing;
2. repair it without the teacher supplying the exact change;
3. use the improved decision on different content;
4. retrieve it after a delay;
5. maintain it during a complete timed composition;
6. repeat that control across more than one unseen sample.

Not every skill needs perfection before the learner moves on. But a single corrected script should never be the only evidence.

The two-success rule

For classroom training, a practical minimum is two independent successful transfer samples before reducing attention to a repair target.

This is not an official Cambridge rule. It is a conservative training heuristic.

If the learner succeeds once and fails once, keep the target active.

If the learner succeeds twice under meaningfully different conditions, move it to the watch list and continue sampling.

The relapse rule

A previously stable weakness can return under:

• harder prompts;
• greater time pressure;
• more complex syntax;
• unfamiliar settings;
• fatigue;
• larger narrative scope.

Relapse does not mean earlier training was wasted.

It reveals the boundary of current control.

Return briefly to the smallest repair that restores the mechanism, then retest under the condition that caused the relapse.

Build a feedback portfolio, not a pile of marked essays

A useful portfolio does not simply store scripts chronologically.

For each composition, keep:

• prompt and date;
• conditions;
• assessment profile if available;
• first weak link;
• preserved strength;
• bounded repair;
• fresh transfer task;
• transfer result;
• current status: now / next / watch / stable.

This makes recurrence visible.

You can see whether “scope inflation” appears every third narrative, whether descriptive attention improves only in familiar places, or whether punctuation fails mainly under timed load.

The feedback heat map

Across six to eight scripts, create a simple table with recurring mechanisms.

Rows might include:

• task fit;
• scope;
• causal development;
• descriptive route;
• viewpoint;
• ending control;
• lexical precision;
• sentence boundaries;
• tense;
• late accuracy drift;
• time allocation;
• cue dependency.

Columns are compositions.

Mark each cell:

S secure, U unstable, N not yet controlled, or blank if not meaningfully tested.

The heat map should not become another scoring system. Its purpose is to stop one memorable script from dominating your diagnosis.

Track conditions beside improvement

A learner’s composition record might look like:

Week 1: Narrative — timed, unseen, independent — scope unstable.

Week 1 repair: six-point plan with tutor questions — scope secure under support.

Week 2: Narrative — untimed, unseen, independent — scope secure.

Week 3: Narrative — timed, unseen, independent — scope mostly secure; ending still slightly compressed.

Week 4: Narrative — timed, unseen, independent — scope secure; target moved to watch.

This record tells a learning story that a single mark cannot.

A four-week feedback-to-transfer programme

Week 1 — Diagnosis

Use one complete marked composition.

Tasks:

• preserve the original;
• translate major comments;
• identify rival causes;
• choose one first weak link;
• perform one bounded repair;
• explain the repair.

Do not immediately write three more full essays.

Week 2 — Near transfer

Use short tasks on changed content.

For narrative development: three different causal plans.

For descriptive attention: two settings with different physical routes.

For sentence control: new sentence relationships, then one paragraph.

For scope: several plans with event deletion.

Reduce teacher explanation.

Week 3 — Farther transfer

Use a substantially different unseen prompt.

The teacher may give only a general target such as “development” or “control”, or no target at all depending on readiness.

Write a complete composition under increasingly realistic conditions.

Afterwards, compare the feedback target before comparing the total mark.

Week 4 — Stability test

Give a new unseen timed composition without specific reminders.

Do not select the prompt to favour the trained skill.

Assess:

• did the repaired mechanism appear independently?
• did it survive full-task load?
• did another weakness become the new first weak link?
• is the original target stable enough to move to watch?

Training advances when the limiting mechanism changes.

Narrative and descriptive transfer should cross-check each other where appropriate

Some mechanisms are genre-specific. Causal plot development belongs mainly to narrative. Controlled focus and overall picture belong strongly to description.

Other mechanisms cross genres:

• proportion;
• sentence control;
• lexical precision;
• viewpoint;
• time management;
• editing;
• independence from prompts and models.

If sentence-boundary control improves only in narrative but fails in description, the grammar repair is not yet fully generalised.

Use marks carefully when tracking feedback

Marks can help when conditions are comparable.

They mislead when:

• one script is revised and another is first-draft;
• one is timed and another untimed;
• one prompt is familiar and another unfamiliar;
• one received tutor prompting;
• different syllabus-year criteria are being applied;
• one marker is using a different standard.

Track mechanism evidence alongside marks.

Variance after feedback

Suppose a learner’s results after a repair are 34, 27, 35, 28.

The average may look acceptable. The variance says the control is unstable.

Investigate the low samples:

• different genre?
• unfamiliar topic?
• time collapse?
• larger plot?
• no teacher reminder?

The lows often reveal where transfer breaks.

High floor versus high ceiling

Feedback can aim at two different improvements.

Raise the ceiling: help the learner produce more sophisticated performance at their best.

Raise the floor: reduce the chance of collapse on an awkward unseen task.

Examination preparation needs both.

A high-performing learner with large variance may benefit more from reliability work than another layer of stylistic sophistication.

Feedback before mock examinations

As the examination approaches, feedback should become lighter and more transferable.

Early phase:

detailed diagnosis, modelling and bounded repair.

Middle phase:

questions and partial cues.

Late phase:

post-task feedback, minimal or no during-task prompting, greater emphasis on execution and recovery.

The learner should not enter the examination needing a tutor’s voice to initiate every strategy.

Feedback after mock examinations

Do not attempt to repair every weakness revealed by a mock.

Identify:

1. which weakness cost the most across the whole response;
2. whether it is a knowledge, selection or execution problem;
3. whether it has appeared before;
4. what can realistically change before the next simulation.

Then choose the next cycle.

AI as a feedback second reader

AI can be useful when its job is constrained.

Useful jobs include:

• locate repeated sentence-boundary patterns;
• compare two possible diagnoses;
• identify whether two paragraphs perform the same job;
• generate a substantially different transfer prompt;
• ask the learner questions rather than rewrite the prose;
• check whether a feedback justification points to actual textual evidence.

Unsafe or misleading uses include:

• “Mark this exactly like a Cambridge examiner and tell me my official grade”;
• trusting criteria the tool has not been given or verified;
• accepting a polished AI rewrite as evidence of student improvement;
• mixing syllabus years;
• treating an AI score as examiner standardisation.

AI can support analysis. It should not impersonate official assessment authority.

A safer AI feedback workflow

1. Provide the correct examination-year criteria or link to the official source.
2. Ask for evidence, not an official grade.
3. Ask for at least two possible diagnoses.
4. Ask which evidence would distinguish them.
5. Have the learner choose and explain the repair.
6. Generate a fresh transfer task with different surface content.
7. Keep the AI output separate from the student’s original script.

The learner remains responsible for the writing decision.

Privacy and student scripts

Before sending student work to any digital service, follow the applicable school, family and platform privacy rules.

Remove unnecessary identifying information.

Do not upload confidential school records, personal identifiers or sensitive material merely to obtain writing feedback.

The educational need is usually the text and task context, not the learner’s identity.

Frequently asked questions about IGCSE composition feedback

Should I rewrite the entire composition after receiving feedback?

Not automatically. If the main weakness is local, a bounded repair may teach more efficiently. Repair enough of the original to understand the mechanism, explain the change, then move to a fresh prompt. A full rewrite is useful when the whole structure must be redesigned or when the learner needs to practise integrating several repairs.

How many teacher comments should I work on at once?

Usually one first weak link plus, at most, one secondary recurring language pattern is enough for a focused cycle. Other comments can enter the next or watch queue. The exact amount depends on the learner, but more comments do not automatically produce more learning.

Why do I keep making the same mistake after correcting it?

Because correction and independent retrieval are different tasks. You may understand the error once it is pointed out but fail to notice or prevent it while managing a new composition. Move from supported repair to uncued production gradually.

How soon should I test transfer?

Use both immediate and delayed tests. An immediate different-surface task confirms that the learner can apply the principle. A delayed task several days later gives stronger evidence that the repair can be retrieved rather than merely held in working memory.

What if my redraft is much better but my next composition is not?

That means the feedback improved revision competence but has not yet transferred reliably. Reduce cues, change the surface further and practise retrieval. Do not simply polish the original redraft again.

Should my teacher correct every grammar error?

Not necessarily. For a repeated pattern, marking representative examples and asking the learner to locate the remaining instances can produce more active learning. Severe errors that block meaning or reflect a new concept may need direct teaching.

Is “show, don’t tell” good feedback?

Only when translated into a specific decision. Some information should be told efficiently. Ask which moment deserves observable detail because it carries character, pressure or perception. A blanket ban on telling can make examination writing bloated.

My teacher says my vocabulary is simple. Should I memorise advanced words?

First identify the mechanism. You may need more lexical range, but you may instead be using vague verbs, repeating familiar adjectives, selecting words imprecisely or failing to retrieve known vocabulary under time. The repair should match the evidence.

Should I use my mark as the main measure of whether feedback worked?

No. Marks are useful when task, conditions, marking standard and syllabus cycle are comparable. Also track whether the specific repaired mechanism appears on different unseen tasks without prompting.

Can my mark go down even though the feedback is working?

Yes. A fresh prompt may be harder. You may be writing more independently, attempting a less familiar genre or expanding syntactic range. Look at the trained mechanism and the conditions before concluding that learning has reversed.

How do I know when a weakness is fixed?

Look for repeated independent success: the learner can recognise the issue, repair it, reproduce the improved decision on changed content, retrieve it after delay and maintain it under full timed composition load. One corrected redraft is not enough.

Should I compare my revised mark with my original mark?

You can, but label the conditions. The difference shows what improvement is possible with feedback and revision. It does not measure independent examination readiness. A later fresh timed script is required for that.

Can peer feedback help?

Yes, especially when peers are asked to point to evidence rather than assign unsupported marks. Useful peer questions include: “Which paragraph changes the story most?”, “Where does the observer move?”, and “Which sentence pattern repeats as an error?”

What if two teachers give different feedback?

Compare the evidence and the level of diagnosis. One teacher may be describing a symptom while another names an upstream cause. If the comments genuinely conflict, return to the script, the correct examination-year criteria and, where available, a shared marking standard. Do not resolve the difference by choosing the comment you prefer.

Should I use model answers after feedback?

Only if the model has a clear teaching job. A model can show what causal development, controlled focus or sentence precision looks like. Close it before the transfer task. The model should clarify a mechanism, not supply a replacement composition.

Can AI give me IGCSE composition feedback?

AI can help locate evidence, propose rival diagnoses, generate fresh prompts or ask useful questions. It should not be treated as an official Cambridge examiner. Supply the correct current criteria, verify factual claims and keep AI-generated rewrites separate from evidence of your own writing ability.

Should I use AI to rewrite my composition?

Not if the purpose is to measure your independent improvement. A rewrite can be studied as a comparison, but it does not prove that you can generate the same decisions yourself. Ask the tool to explain or question before asking it to replace your prose.

How close to the exam should I still be rewriting old essays?

As the examination approaches, shift toward fresh unseen execution and lighter post-task diagnosis. Old scripts remain useful for reviewing recurring patterns, but examination readiness depends on independent performance under realistic conditions.

What should I do the day after receiving feedback?

Complete one small retrieval task without the corrected sentence or paragraph visible. Then schedule a changed-surface task. The day after feedback is a good moment to discover whether you remember the principle or only recognise the old correction.

What should I do if I receive no detailed teacher feedback?

Use the official criteria for your examination year, preserve the script and perform an evidence-led self-review. Compare your judgement with a teacher or tutor when possible. Start with one observable pattern rather than inventing an exact official mark.

Does every feedback cycle need a full mark?

No. Many useful cycles should focus on one mechanism without scoring the entire piece. Constant scoring can distract from learning. Use full mark-scheme judgement when the complete performance is being sampled.

Version control: feedback must stay attached to the examination year

Cambridge IGCSE First Language English 0500 has separate current syllabus cycles. Candidates sitting in 2026 should use the 2024–2026 documents. Candidates sitting from 2027 should use the 2027–2029 syllabus and relevant assessment materials.

The feedback mechanics in this guide are intentionally durable. The exact descriptors that generated a teacher’s comment may not be.

Write the examination year on major feedback records.

Example:

0500 / 2027–2029 / Paper 2 Composition / Content and Structure development.

That small label prevents future confusion when resources from different years accumulate.

Do not transfer the mark scheme across qualifications automatically

This page is specifically an IGCSE First Language English composition-feedback adapter. Similar writing problems occur in GCSE, Pearson International GCSE, IB, AP, VCE, NCEA and other systems, but the assessment architecture differs.

Transfer the writing mechanism only after checking the new qualification’s task and criteria.

For example:

“sentence boundaries collapse under complexity” is a general writing mechanism.

“this exact Cambridge Style and Accuracy level applies” is not.

Canonical ownership inside SETC

This page owns a narrow but important job:

marked Cambridge IGCSE composition → evidence → diagnosis → bounded repair → fresh unseen transfer.

It does not own every neighbouring job.

Mark-scheme judgement

Use IGCSE Composition Mark Scheme | Judge a Script with Evidence, Not Guesswork when the main question is how to judge the completed script, choose a best-fit level, moderate disagreement or self-assess with the appropriate criteria.

Content and Structure teaching

Use IGCSE Composition Content and Structure when the learner needs the full system for development, progression and coherence rather than a feedback workflow.

Style and Accuracy teaching

Use IGCSE Composition Style and Accuracy for the deeper language-control owner.

Recovery during writing

Use IGCSE Composition Recovery when the draft is already failing under examination pressure.

Error diagnosis and plateau

Use IGCSE Composition Mistakes for the broader failure taxonomy.

Practice progression

Use IGCSE Composition Practice System for the full untimed-to-unseen training progression.

Model-answer study

Use IGCSE Composition Model Answers when the specific problem is how to study exemplars without memorising them.

Generic English feedback theory

SETC already has the broader English Exam Feedback Transfer | Turn Corrections into Better Future Performance owner. This IGCSE page does not replace it. It applies that broader principle to Cambridge composition with genre-specific case files, examination-year boundaries and IGCSE composition transfer probes.

Where the wider eduKate ecosystem fits

The wider eduKate estate contains specialised neighbouring material, including Cambridge IGCSE vocabulary work. Those owners can supply lexical depth without being copied into this page.

SETC’s job here is orchestration: when vocabulary feedback appears inside a complete IGCSE composition, determine whether the issue is knowledge, precision, selection, retrieval or forced sophistication, then route deeper lexical training to its owner.

A mature ecosystem becomes more useful when each page knows what not to reteach.

Turn feedback into better performance on an unseen task

A common instruction is:

“Read your teacher’s comments, correct your mistakes, rewrite the essay and practise again.”

That advice is not wrong. It skips almost every hard decision.

The hard reader job includes:

• distinguishing comment from mechanism;
• separating correction from transfer;
• testing rival diagnoses;
• finding the first weak link;
• choosing the minimum useful repair;
• preserving existing strengths;
• working across narrative and descriptive feedback;
• handling sentence-control errors without flattening ideas;
• recognising when an ending problem is really a scope problem;
• fading teacher cues;
• measuring transfer across surface, cue, delay and task load;
• tracking false progress;
• using marks only under comparable conditions;
• building a feedback portfolio;
• moving from redraft competence to unseen stability.

Those operations belong together because the reader begins with one marked script and needs to arrive at a different future performance.

Check progress by asking whether the learner can apply the repaired decision on a new task, after a delay and without the original feedback beside them.

The final feedback standard

Maren holds up a marked paragraph.

“I fixed it.”

Iona reads the revision.

“Yes. What was wrong?”

“I kept telling the reader the atmosphere instead of choosing evidence.”

“What caused it?”

“I didn’t know where the observer was looking, so I used general labels.”

Leonie closes the old composition.

“Good. Now prove it somewhere else.”

That is the standard.

Feedback begins on an old script.

Learning ends on a new one.

Official examination references

Always verify the current assessment documents for the candidate’s examination year. Start with the official Cambridge IGCSE First Language English 0500 subject page.

For examinations from 2027, use the official 2027 specimen Paper 2 and 2027 specimen Paper 2 mark scheme alongside the 2027–2029 syllabus.

Candidates sitting in 2026 should use the 2024–2026 syllabus and the official materials applicable to that cycle.

Feedback drill library — practise the repair without rewriting another full composition

The drills below are deliberately smaller than a complete IGCSE composition. Their purpose is to isolate one mechanism after feedback, increase difficulty gradually and then return the learner to full unseen writing.

Drill 1 — Paragraph-job duplication

Take a narrative with six paragraphs.

Write a five-word job beside each paragraph.

If two consecutive paragraphs have the same job, choose one:

• delete one;
• combine them;
• make the second change pressure, information or choice.

Then apply the same test to a fresh plan.

Drill 2 — Consequence chain

Write five linked statements beginning with one small choice.

Example frame:

Because the character chooses ___, ___ becomes possible.

Because that happens, the character can no longer ___.

This forces the character to choose between ___ and ___.

The choice causes ___.

The ending changes ___.

Repeat with a different setting and relationship.

Drill 3 — Remove one coincidence

Find an event in a narrative that arrives from outside the character’s choices.

Ask:

Can this event become a consequence of something already established?

If yes, rewrite the causal link.

If not, decide whether the event is necessary.

Drill 4 — The three-event ceiling

Plan a narrative using only three major events after the opening.

You may add minor actions, but no fourth major event.

Use the saved space for motive, choice and consequence.

This drill is useful for students whose feedback repeatedly says “too much happens”.

Drill 5 — Decision before drama

Write a climax with no accident, chase, injury, shouting match or surprise villain.

The central pressure must turn through:

• a refusal;
• an admission;
• a choice to leave;
• a choice to stay;
• a handover;
• a deletion;
• a signature;
• a silence that has consequence.

The aim is to learn that consequence, not spectacle, creates narrative intensity.

Drill 6 — Ending-state contrast

Write two sentences:

Opening state: what does the character believe, want or avoid?

Ending state: what has changed?

If the two sentences are nearly identical, the story may lack development.

Build only the middle needed to make the change credible.

Drill 7 — Atmosphere-label ban

Describe a place without using:

tense, peaceful, chaotic, scary, beautiful, sad, exciting, gloomy, magical, busy, noisy or quiet.

The reader must infer the effect from observable detail.

Drill 8 — Fixed observer

Choose one physical position.

For 180 words, the observer cannot move.

Structure the paragraph through changes of attention, distance, sound, light or behaviour rather than walking through the place.

Drill 9 — Moving observer

Now do the opposite.

Give the observer a physically credible route through a place.

Each stop must perform a different descriptive job.

If two stops reveal the same thing, remove one.

Drill 10 — Behaviour creates weather

Describe approaching rain, heat, wind or cold without describing the sky for the first 120 words.

Show the weather through:

• what people prepare;
• how objects move;
• what changes direction;
• what becomes difficult;
• what disappears or appears.

Drill 11 — Detail hierarchy

Write ten details for a setting.

Rank them from most to least useful for the chosen effect.

Use only the top four.

This trains selection rather than accumulation.

Drill 12 — Repeat with changed meaning

Choose one ordinary object.

Write one sentence showing it in the opening.

Then design a change in the scene or character.

Write a final sentence where the same object now carries different significance.

If the meaning has not changed, repetition is decorative rather than structural.

Drill 13 — Evidence before adjective

Choose five adjectives from a draft.

For each, write one observable piece of evidence that would allow a reader to infer the quality.

Then decide whether the adjective is still needed.

Drill 14 — Vague verb repair

Find went, got, did, looked, moved or made in a draft.

Do not replace them with rarer synonyms automatically.

Ask what actually happened.

Replace only when a more precise verb carries useful information.

Drill 15 — Collocation check

Select five “advanced” words from a composition.

Check each in a reliable dictionary or corpus-supported reference.

Record:

• meaning;
• usual grammatical pattern;
• common collocations;
• register;
• whether the original sentence remains natural.

Remove any word chosen mainly for rarity.

Drill 16 — Two ways to express cause

Write one cause-and-effect idea.

Express it in two grammatically correct ways without changing the meaning.

Example relationships:

• because;
• so;
• therefore in a separate sentence;
• cause first, consequence second;
• consequence first, cause in a subordinate clause.

The goal is choice, not a preferred template.

Drill 17 — Comma-splice hunt

Take a timed draft.

Underline every comma.

For each, ask whether the words on both sides could stand as complete sentences.

If yes, check whether the punctuation and conjunction are grammatically justified.

Then write three fresh sentences using the repaired relationship.

Drill 18 — Dialogue boundary control

Write six lines containing:

• speech followed by a reporting clause;
• reporting clause followed by speech;
• speech, action, then separate speech;
• a question in dialogue;
• an interruption;
• speech without a reporting clause.

Check punctuation, then use only two of the patterns in a fresh narrative paragraph.

Drill 19 — Tense stability

Take a paragraph written in past tense.

Circle every finite verb.

Explain any tense change.

If the change has no clear time or viewpoint reason, repair it.

Then write a new paragraph with one deliberate earlier-time reference and return cleanly to the main narrative tense.

Drill 20 — Pronoun-reference test

Find every he, she, they, it, this and that in one paragraph.

Ask whether the reader can identify the reference instantly.

Repair ambiguity without replacing every pronoun with a noun.

Drill 21 — Syntax under load

Write one accurate paragraph using mostly simple and compound sentences.

Then revise only two sentences to express more complex relationships.

If accuracy collapses, the training target is those relationships—not “write more complex sentences everywhere”.

Drill 22 — The late-script accuracy test

Compare the first 100 words and last 100 words of a timed composition.

Count patterns, not just errors:

• sentence-boundary failures;
• spelling slips;
• vague words;
• unfinished constructions;
• missing punctuation.

If the last 100 words deteriorate repeatedly across scripts, investigate execution and scope.

Drill 23 — Cut twenty per cent

Take a narrative plan with ten bullet points.

Delete two without damaging the central pressure.

Then delete another if possible.

Explain why the remaining events are load-bearing.

Drill 24 — Climax checkpoint

During practice, note when you reach the decisive turn.

Do this across three timed compositions.

If the climax repeatedly arrives near the end of available writing time, reduce scope in the next plan.

Drill 25 — Preserve one strength

Before revision, name one feature you will not damage.

Examples:

• concise opening;
• two-character scope;
• clear viewpoint;
• accurate simple sentences;
• specific observational detail.

After revision, verify that the strength remains.

Drill 26 — Rival diagnosis

Choose one visible weakness.

Write two plausible causes.

Then write one piece of evidence you would expect if each cause were correct.

Use another script or another condition to test which cause is stronger.

Drill 27 — Comment translation

Take one vague comment:

“awkward”, “develop”, “structure”, “vocabulary”, “be descriptive”, “check grammar”.

Rewrite it as:

Evidence → mechanism → repair → transfer test.

If you cannot fill all four, ask for clarification or gather more evidence.

Drill 28 — Teacher-cue fade

Perform the same repair across four tasks:

1. teacher tells you the exact problem;
2. teacher gives a question;
3. teacher gives only a broad target;
4. teacher gives no cue.

Record where performance breaks.

Drill 29 — Fresh-surface generator

After repairing one model or script, change at least four variables:

• setting;
• relationship;
• central object;
• time of day;
• viewpoint;
• type of pressure;
• ending state.

Then use only the underlying principle.

Drill 30 — Twenty-four-hour retrieval

The day after feedback, write the mechanism from memory in one sentence.

Do not reread the teacher’s comment first.

Create a five-minute practice task that uses it.

Only then check the original feedback.

Drill 31 — One-week stability probe

One week later, include the same mechanism inside a full unseen task without announcing that it is being tested.

After writing, inspect whether the control appeared naturally.

Drill 32 — Description-to-narrative language transfer

If lexical precision has improved in description, test it inside narrative action.

A precise vocabulary system should not disappear when plot pressure increases.

Drill 33 — Narrative-to-description sentence transfer

If sentence boundaries are secure in dialogue and action, test them in a descriptive paragraph with longer spatial relationships.

The grammatical repair should survive genre change where appropriate.

Drill 34 — No-score feedback session

Complete one composition and forbid yourself from calculating a total mark.

Identify:

• one preserved strength;
• one first weak link;
• one bounded repair;
• one transfer task.

This breaks the habit of treating every practice as a score event.

Drill 35 — Profile comparison

Take two compositions with similar total marks.

Compare why they reached those totals.

One may have stronger Content and Structure; the other stronger Style and Accuracy.

Give each a different next task.

Drill 36 — The support-removal test

Repeat a successful practice without:

• the model;
• the vocabulary bank;
• the paragraph frame;
• the tutor’s pre-task reminder.

Remove only the support you are ready to test.

Success under reduced support is stronger evidence of independence.

Drill 37 — The “what survives?” reflection

After every full feedback cycle, answer:

1. What did I change?
2. What did I preserve?
3. What transferred?
4. What still depends on a cue?
5. What is the next first weak link?

Keep the answers brief enough to read before the next cycle.

Drill 38 — Transfer without the feedback vocabulary

If the teacher used the word “causality”, write the next task without that word in the instructions.

If the teacher used “viewpoint”, remove the label.

If the teacher used “sentence boundaries”, remove the reminder.

The learner should eventually retrieve the decision from the task, not from the terminology.

Drill 39 — New first weak link

Once a repair is stable, deliberately stop making it the centre of every lesson.

Write a fresh composition and identify what now limits the performance most.

Progress often looks like the appearance of a more advanced problem.

Drill 40 — Final examination simulation

Choose a fresh task under the correct conditions for your syllabus year.

No model. No teacher prompts. No feedback card during writing.

After completion:

1. preserve the script;
2. judge it with the correct assessment criteria;
3. check whether previously trained mechanisms remained stable;
4. record the new first weak link;
5. decide whether another repair cycle is needed.

The feedback system has succeeded when the learner can perform this cycle with decreasing external help.

How to use the drill library without creating worksheet dependency

Do not complete all forty drills.

Select the drill that matches the diagnosed mechanism.

Use it briefly.

Increase transfer distance.

Return to whole composition.

The drill is a bridge back to authentic performance.

If the learner becomes excellent at the drill but cannot use the skill in a complete unseen task, the bridge has not yet reached the other side.

Continue with English Writing and Composition | Give Ideas Shape and Purpose for the connected guides and the next practice route.