A continuous-writing response can be fluent sentence by sentence and still feel unreliable because the narrator suddenly knows something they could not have seen, heard or learned. This is a viewpoint-knowledge leak. The repair is not simply to keep using “I” or “she”. You must control the evidence available to the narrator at each moment.
Use a knowledge boundary: for every important claim, ask whether the narrator observed it, heard it, inferred it from visible clues, remembered it, or learned it later. If none applies, either provide a route by which the information becomes available, change the claim into a reasonable inference, or move the narration to a viewpoint that can legitimately know it.
This G2 English workshop is a narrow application within continuous writing. It adds practice beyond the earlier owner on deixis and reference by focusing on access to knowledge, not merely who “I”, “here” and “now” denote. All examples are original and are not examination questions.
Find your next learning step
Define the boundary (chapters 1–5) · Control narrative access (chapters 6–11) · Diagnose and repair leaks (chapters 12–17) · Practise and transfer (chapters 18–27)
1. Locate the skill inside K200 English
2. Viewpoint is more than first or third person
3. Build the narrator’s evidence channels
4. Knowledge changes across time
5. Separate fact, inference and suspicion
6. First-person limits can create strength
7. Third-person limited needs a centre
8. Omniscience must be deliberate, not accidental
9. Dialogue transfers knowledge but not automatically truth
10. Internal reactions must belong to the viewpoint
11. Place and distance constrain knowledge
12. Worked scene: identify the first leak
13. Diagnose whether the leak is necessary
14. Repair method one: change certainty
15. Repair method two: change the evidence channel
16. Repair method three: move or change the narrator
17. Control revelation order
18. A three-pass editing routine
19. Practice ladder
20. Explained practice
21. For parents and tutors
22. Final control: let knowledge arrive honestly
23. Workshop: the same event from two knowledge positions
24. Workshop: fair surprise rather than hidden explanation
25. Workshop: viewpoint under time pressure
26. Transfer beyond narrative
27. Compact mastery check
1. Locate the skill inside K200 English
The 2027 K200 syllabus describes continuous writing in Paper 1 Section C. Candidates write 250–400 words on one of four topics, and possible text types include narrative or argument. This guide addresses a craft problem most directly relevant to narrative writing: keeping information consistent with the chosen viewpoint.
The official descriptors value coherent presentation, clarity and accurate language. A knowledge leak can damage coherence because the reader no longer understands how information reached the narrator. The issue is therefore not a secret examiner rule; it is a practical way of protecting meaning and organisation.
The exact topic and wording remain in control. Do not force a narrative when another response type better answers the selected task. When you do choose narrative writing, the viewpoint should support the story rather than become a source of accidental contradictions.
During practice, label the chosen viewpoint at the top of your planning page. Then list the important facts the narrator can know at the opening. This small map prevents later scenes from borrowing information that belongs only to another character.
Back to contents · Previous chapter · Next chapter
2. Viewpoint is more than first or third person
First person uses a narrator who speaks as “I”. Third person uses names or pronouns such as “she”, “he” or “they”. But grammatical person alone does not determine knowledge. A first-person narrator can report only their own access, while third-person narration may be limited to one character or range more widely if deliberately controlled.
A story can keep every pronoun correct and still leak knowledge. “I watched Mei close the envelope. Across town, she wondered whether I had guessed the secret.” The narrator cannot directly know Mei’s private thought at that moment unless the narrative establishes how it was learned.
The safer first-person version reports evidence: “Across town, Mei may have wondered whether I had guessed; when she called later, her voice kept breaking.” The inference is marked as inference, and the later call supplies observable support.
When reviewing, separate pronoun accuracy from access accuracy. Both matter, but they solve different problems.
Back to contents · Previous chapter · Next chapter
3. Build the narrator’s evidence channels
A narrator can know through direct observation, spoken or written communication, memory, legitimate background knowledge and reasonable inference. Each channel leaves a trace in the writing. The reader sees the gesture, hears the sentence, reads the message or recognises the clue that supports the conclusion.
Direct observation is strongest for visible actions and sounds: the lock clicked, the light went out, the teacher paused. It does not automatically reveal motives. “He paused because he hated me” converts an action into certainty without a route.
A calibrated alternative is “He paused, and I took it as another sign that he did not trust me.” Now the hostility is the narrator’s interpretation. The story may later confirm or overturn it.
Create a five-column planning row: fact, who knows, how learned, when learned, certainty. Use it only for decisive information; the aim is control, not paperwork.
Back to contents · Previous chapter · Next chapter
4. Knowledge changes across time
A narrator may learn later what was unknown during the event. Retrospective narration can reveal this, but the time shift must be clear. “I did not know then that the message had been sent from my brother’s phone” distinguishes past ignorance from later knowledge.
Without the phrase “I did not know then”, the same sentence may imply impossible awareness. The solution is not always to remove the information. Often you need a temporal bridge showing when and how the narrator acquired it.
Use three checkpoints: before the scene, during the scene and after the scene. Place each revelation at the earliest point when the narrator can legitimately possess it. This also helps suspense, because important facts arrive in a controlled order.
Do not overuse hindsight announcements. If every paragraph promises a future revelation, the device becomes mechanical. Reserve it for information whose later acquisition genuinely changes the meaning of the earlier scene.
Back to contents · Previous chapter · Next chapter
5. Separate fact, inference and suspicion
Narrative certainty has levels. “The drawer was open” can be observed. “Someone had searched it” is an inference from the state of the drawer. “Ravi searched it” is a stronger claim requiring identifying evidence. “Ravi wanted me expelled” adds motive and needs even more support.
Many knowledge leaks occur because the sentence jumps two levels at once. The learner sees an open drawer and writes a complete explanation of another character’s plan. The prose may sound decisive, but it removes credibility.
Repair by choosing the strongest claim supported at that moment. You can preserve tension with calibrated language: perhaps, must have, seemed, I suspected, only later did I learn. These expressions should reflect real uncertainty rather than decorate the paragraph.
Practice on one scene by writing four versions: observation, inference, accusation and later confirmation. Notice how the required evidence changes.
Back to contents · Previous chapter · Next chapter
6. First-person limits can create strength
A first-person narrator’s limitation is not a weakness. It can create discovery, misinterpretation and surprise. The reader learns with the narrator and can evaluate the narrator’s judgments. The important condition is that uncertainty remains honest.
Suppose the narrator hears two people stop talking when she enters. She may feel excluded and suspect that they discussed her. She cannot report their exact earlier conversation unless she later receives it through a message, recording or confession.
Use sensory evidence to keep the scene active: the unfinished sentence, the phone turned face down, the glance exchanged. These details allow the reader to form an inference without granting the narrator impossible access.
Then decide whether the suspicion is correct. A later reversal works only if the initial evidence supported more than one plausible reading.
Back to contents · Previous chapter · Next chapter
7. Third-person limited needs a centre
Third-person limited narration usually follows one viewpoint character closely. The prose can enter that character’s thoughts but treats other minds from the outside. Establish the centre early through perception, concern and selection of detail.
If Aisha is the centre, “Aisha counted the empty chairs and wondered who had left” is inside her access. “Mr Lim concealed his relief” may be acceptable if relief is visibly inferred, but “Mr Lim remembered the argument and decided to punish her” enters his mind without a deliberate shift.
Repair by externalising: “Mr Lim released a long breath, then folded the complaint into his pocket.” Or begin a clearly separated new scene centred on him if the structure truly needs his private decision.
During editing, circle every private thought. Check that all circles belong to the same centre within the scene.
Back to contents · Previous chapter · Next chapter
8. Omniscience must be deliberate, not accidental
An omniscient narrator may know more than any one character, but successful use requires stable control. Rapidly entering several minds can fragment a short examination piece and consume words explaining transitions. Accidental omniscience usually feels like a leak rather than a design.
For a 250–400 word task, a single controlled viewpoint is often easier to develop coherently. That is a practical choice, not an examiner requirement. A capable learner may use a broader narrator when the topic and plan justify it.
Test necessity: does entering the second mind reveal something the story cannot show through action, dialogue or later discovery? If not, keep the original centre and protect continuity.
The advanced skill is not avoiding complexity at all costs. It is selecting complexity that earns its place within the available length.
Back to contents · Previous chapter · Next chapter
9. Dialogue transfers knowledge but not automatically truth
When one character tells another something, the listener gains access to the statement, not guaranteed access to the truth. “Nadia said the meeting was cancelled” is different from “The meeting was cancelled.” The first records communication; the second asserts the event.
This distinction creates useful narrative possibilities. A character can lie, misunderstand or repeat a rumour. The narrator can believe the statement while the story leaves room for correction.
Mark the source with verbs and context: said, claimed, admitted, warned, wrote. Avoid choosing a loaded verb unless the narrator can justify it. “Confessed” presumes wrongdoing; “said” may be safer before the facts are established.
Practice rewriting one information transfer in three ways: trustworthy report, uncertain claim and deliberate deception. Make the evidence visible rather than explaining the label.
Back to contents · Previous chapter · Next chapter
10. Internal reactions must belong to the viewpoint
A viewpoint character’s bodily sensations, memories and thoughts can be narrated directly. Another character’s inner state must normally be inferred from observable signs. “My stomach tightened” is direct first-person access; “her stomach tightened” requires third-person access centred on her or an external description.
External signs include posture, movement, speech and timing. “She answered too quickly and kept rubbing the edge of the ticket” permits anxiety as an inference. It does not prove the exact cause.
Avoid translating every gesture into a named emotion. Sometimes the detail itself is more vivid and allows the reader to participate. Use explicit emotion when it clarifies the viewpoint character’s experience or moves the scene.
A useful editing question is: could a camera or microphone capture this? If yes, an outside observer can report it. If no, identify whose inner access makes it available.
Back to contents · Previous chapter · Next chapter
11. Place and distance constrain knowledge
Characters cannot normally observe events in another room, building or time without a channel. Phones, messages, windows, recordings and later reports can bridge distance, but the channel must exist before its information appears.
A common leak is the cutaway sentence: “While I waited outside, the panel decided to reject me.” The first-person narrator cannot know the decision during the wait. Repair it as later knowledge: “I learned afterwards that the panel had already decided.”
Distance can strengthen suspense. Keep the narrator outside, let them interpret muffled voices and delayed movement, then reveal the decision through the door opening or an official message.
Draw a simple location line for complex scenes. If information crosses it, mark the channel. This takes seconds and prevents impossible seeing or hearing.
Back to contents · Previous chapter · Next chapter
12. Worked scene: identify the first leak
Original draft: “I left the classroom after placing the model on the table. Behind me, Marcus opened my bag, found the missing component and smiled because his plan was working. At the bus stop, I wondered why everyone avoided me.” The first sentence is within the narrator’s access. The next sentence is not.
The earliest leak is Marcus opening the bag after the narrator has left, unless the narrator observes it through some established channel. His discovery and private motive deepen the breach. The bus-stop thought is valid first-person access.
Repair A keeps the viewpoint: “When I returned, my bag stood open and the missing component lay beside Marcus’s model. He looked away as I approached.” The narrator now has evidence but not certainty about the full plan.
Repair B uses retrospective knowledge: “Only later, after the corridor camera was checked, did I learn what Marcus had done.” This changes the time of access and should be integrated carefully so it does not solve the whole story too easily.
Back to contents · Previous chapter · Next chapter
13. Diagnose whether the leak is necessary
Before repairing, ask what job the leaked information performs. Does it create dramatic irony, explain causation or merely save the writer from staging a discovery? The answer determines the best repair.
If dramatic irony is essential, a deliberately broader narrator or a clear scene shift may work. If the information merely explains too soon, remove it and let later evidence reveal the cause. If the scene needs urgency, transfer the information through a credible message.
Do not preserve every sentence. A clean deletion can improve both viewpoint and pacing. Examination writing benefits when each detail advances the task rather than defending an early draft choice.
Write the job in the margin: reveal, mislead, explain, motivate or foreshadow. Then choose the least disruptive valid route.
Back to contents · Previous chapter · Next chapter
14. Repair method one: change certainty
When the narrator has clues but not proof, convert the claim into an inference. Replace “she had hidden the key” with “the empty hook made me suspect that she had taken the key.” Now evidence and uncertainty are both visible.
The repair must not become vague padding. Name the clue and the interpretation. “Something felt wrong” communicates less than “the seal was broken although I had checked it that morning.”
Later evidence can confirm or challenge the inference. This creates a fair narrative chain: observation, interpretation, action, new evidence, revision. Each stage belongs to what the narrator knows then.
Practise by finding five unsupported certainty verbs in a draft—knew, realised, intended, planned, hated. Ask whose knowledge licenses each one.
Back to contents · Previous chapter · Next chapter
15. Repair method two: change the evidence channel
Sometimes the claim should remain certain because the story needs it. Add a credible channel: a message arrives, a witness reports the event, the narrator overhears a conversation or a document is discovered.
Choose a channel consistent with the setting and the word limit. A long technical explanation of how evidence was obtained can overwhelm the main event. One well-placed line may be sufficient if it is plausible.
Channels have reliability. An official record, a frightened witness and an anonymous message do not carry equal certainty. Let the narrator’s response match the source rather than treating all communication as fact.
After adding a channel, check timing. Information cannot influence an earlier decision if the narrator receives it later unless the narration explicitly looks back.
Back to contents · Previous chapter · Next chapter
16. Repair method three: move or change the narrator
If the story’s central event occurs away from the current narrator, consider moving the scene or selecting a different viewpoint from the beginning. This is a structural repair and is usually easiest during planning, not in the final minute.
Choose the character with the most useful access and the strongest stake in the event. The person who knows everything may not be the most engaging; the person who must discover the truth can create a clearer arc.
Do not change “I” to “she” and assume the problem is solved. Third person still needs a knowledge model. Identify whose perceptions organise each scene.
For a short response, one stable centre plus a later revelation often works better than several abrupt viewpoint switches. Test both versions during untimed practice.
Back to contents · Previous chapter · Next chapter
17. Control revelation order
Knowledge boundaries shape plot. A revelation should arrive when the narrator has a credible route to it and when it changes the next decision. Too early, it removes uncertainty; too late, it may feel withheld artificially.
Plan three revelations: what the narrator believes, what new evidence appears and what they understand afterwards. The final understanding should be traceable to the new evidence, not to sudden authorial explanation.
Example: the narrator believes a friend ignored a message. A delivery failure notice appears. The narrator realises the accusation was premature and chooses to apologise. Each transition is supported.
This chain develops character without entering the friend’s private mind. The friend can later explain their experience through dialogue if the task needs it.
Back to contents · Previous chapter · Next chapter
18. A three-pass editing routine
Pass one circles viewpoint markers: I, she, his, here, then, later. Pass two underlines private states and strong claims. Pass three draws an arrow from each claim to its evidence channel. Missing arrows identify likely leaks.
Do not edit every sentence equally. Focus on turning points, accusations, motives and revelations because they carry the largest knowledge burden. A minor description of weather rarely threatens viewpoint control.
Then read the story as the narrator. At each paragraph, ask “What do I know now that I did not know before?” If the answer appears without an event or channel, repair the transition.
Finish with ordinary language editing. Viewpoint control does not replace clarity, paragraphing, grammar or task fulfilment; it gives those features a coherent information path.
Back to contents · Previous chapter · Next chapter
19. Practice ladder
Stage one: classify isolated sentences as observation, inference, report, memory or unsupported access. Stage two: repair a five-sentence paragraph. Stage three: write a 120-word scene from one viewpoint and reveal one fact credibly.
Stage four: rewrite the same scene from another character’s viewpoint. Notice which facts become available and which disappear. Stage five: complete a full continuous-writing response under an appropriate practice condition.
After each attempt, record the first leak rather than counting only grammar errors. The first leak often shows the point where planning stopped controlling the story.
Retest with a different setting. Transfer is visible when the learner controls knowledge in a new scenario, not when they reproduce the same repaired paragraph.
Back to contents · Previous chapter · Next chapter
20. Explained practice
Sentence: “From the corridor, I knew the principal was reading my letter and regretted doubting me.” Two problems appear. The narrator cannot see the exact action behind a closed door, and cannot directly know the principal’s regret.
Repair: “From the corridor, I watched the shadow pause behind the frosted glass. When the principal opened the door, she held my letter and said, ‘I should have listened earlier.’” The visual clue and dialogue provide channels.
Alternative repair: “From the corridor, I imagined the principal reading my letter.” This protects access but keeps the claim uncertain. It is suitable if the later scene is meant to reveal a different outcome.
The best version depends on the narrative job. The skill is not choosing one formula; it is making knowledge and certainty match the available evidence.
Back to contents · Previous chapter · Next chapter
21. For parents and tutors
Do not mark every viewpoint leak as “wrong tense” or “confusing”. Quote the exact claim and ask how the narrator knows it. This question lets the learner diagnose access rather than guess at a stylistic preference.
Use colour sparingly: one colour for direct evidence, one for inference and one for later report. If a decisive sentence has no colour, discuss its route. The visual map should reveal thinking, not decorate the page.
Separate ambitious design from accidental switching. A learner may deliberately experiment with multiple viewpoints. Ask them to explain the transition and purpose. If they cannot, stabilise one centre before adding complexity.
Praise successful restraint. A sentence that admits uncertainty can be more precise and more compelling than an unsupported dramatic claim.
Back to contents · Previous chapter · Next chapter
22. Final control: let knowledge arrive honestly
Choose a viewpoint, define its starting knowledge and let new information arrive through visible channels. Distinguish what the narrator observes, infers, suspects, remembers and learns later. That structure keeps the reader oriented.
When a leak appears, change the certainty, add a credible channel, move the revelation or redesign the viewpoint. Do not rely on pronoun consistency alone.
The result is not merely technical correctness. Controlled access can create suspense, misunderstanding, discovery and growth because the reader receives information in a meaningful order.
Return to task fulfilment at the end. A coherent viewpoint must still serve the chosen K200 topic, develop the relevant content and remain clear within the required response.
Back to contents · Previous chapter · Next chapter
23. Workshop: the same event from two knowledge positions
A scholarship letter disappears before an assembly. Version A follows the student who expected it. Version B follows the office assistant who misplaced it. The physical event is the same, but available knowledge differs from the opening sentence.
In Version A, the student sees an empty noticeboard, hears whispers and notices the assistant searching a drawer. The student may suspect rejection or error but cannot know the letter was misfiled. Tension comes from limited access.
In Version B, the assistant remembers placing the envelope under another folder and sees the student waiting. The assistant can know the cause but may not know the student’s private fear. The emotional focus changes even when the plot does not.
Write 120 words from each position and highlight facts available in only one version. Viewpoint is an information architecture, not merely a pronoun choice.
Back to contents · Previous chapter · Next chapter
24. Workshop: fair surprise rather than hidden explanation
A fair surprise rearranges evidence already seen or introduces new evidence through a credible channel. An unfair surprise depends on the narrator secretly knowing a decisive fact while withholding it without reason.
Suppose a narrator accuses a friend of taking a key. Earlier details show a wet footprint and a broken plant pot. A later discovery that a neighbour entered through the balcony can be fair because the clues support an alternative route.
It becomes unfair if the narrator had watched the neighbour enter but never mentioned it while claiming complete uncertainty. The missing fact was concealed by the writer, not unavailable to the narrator.
List the clues the viewpoint character notices and the interpretations they make. The ending may revise the interpretation, but it should not pretend the observations never existed.
Back to contents · Previous chapter · Next chapter
25. Workshop: viewpoint under time pressure
When time is short, protect three points: one centre per scene, no unsupported access to another mind and a visible route for each revelation. These controls prevent the largest coherence failures without elaborate planning.
Use a six-line plan: viewpoint character, place, initial belief, decisive observation, new information channel and changed decision. It is compact enough for practice and specific enough to constrain the draft.
If the draft switches minds, mark the first switch. Decide whether to externalise the other character, move a scene break or keep the new centre. Repair from the first divergence.
In final checking, inspect motives and secret information first. They carry more knowledge risk than ordinary descriptions of clothing or weather.
Back to contents · Previous chapter · Next chapter
26. Transfer beyond narrative
Knowledge-source control also helps comprehension and situational writing. In comprehension, distinguish what the text states from what the reader infers. In a report, distinguish verified information from a recommendation or prediction.
The forms differ, so do not apply narrative devices mechanically. A report needs explicit sourcing and clear claims rather than suspense. The shared principle is that a statement needs a visible basis appropriate to its purpose.
Ask across tasks: who makes the claim, what evidence is available and how certain may the claim be? This creates continuity without collapsing different English components into one template.
Return to the relevant owner for full component teaching. This workshop contributes one transferable discipline: information should arrive through an honest route.
Back to contents · Previous chapter · Next chapter
27. Compact mastery check
Draft: “I saw Lena leave the hall. In the taxi, she deleted the message because she feared I had read it.” The first sentence is available to the narrator. The second enters Lena’s remote action, motive and fear without a channel.
Limited-viewpoint repair: “I saw Lena leave the hall. Minutes later, the message disappeared from our chat. I wondered whether she thought I had read it.” The deletion is observed through the app; motive becomes inference.
Retrospective repair: “Lena later admitted that she had deleted the message in the taxi because she feared I had read it.” Dialogue transfers the information later. The story must place this admission after the uncertainty it resolves.
Mastery is choosing the repair that serves the plot while keeping access honest. The learner should explain not only what changed but how the narrator can now know it.
Back to contents · Previous chapter · Next chapter
Continue your learner route
This workshop extends G2 Vol 0173: English Deixis Control from reference words into narrative access. Use Examination Craft for the wider performance system and eduKateSG English Learning Hub for subject development.
Return to PSLE Learner’s Guide Vol 0080 when the earlier repair is stopping after the complete answer rather than adding unsupported claims.
Continue within this new batch: G2 Vol 0180: Instruction Authority Ladder.
Official references and scope
Framework checked on 13 October 2026: SEAB 2027 G2 syllabuses lists G2 English Language as K200. Read the 2027 K200 G2 English Language syllabus.
K200 Paper 1 Section C requires one continuous-writing response of 250–400 words from four topics. This guide teaches an original narrative-control practice and does not predict topics or impose a compulsory narrative method.