Do Not Repair the Loudest Symptom First
A weak composition may look like a vocabulary problem, an ending problem, a creativity problem or a careless-mistake problem. But the visible failure can appear much later than the point where the writing system first broke.
If the student chose an unworkable route, the ending may collapse. If the student never understood the topic, stronger sentences only make the wrong answer more polished. If language is accurate during drills but falls apart in a full composition, the issue may be load rather than knowledge.
The Control Tower therefore asks:
Where is the earliest weak link that can explain the later failures?
The Composition Diagnostic Chain
Scan in this order:
Task Interpretation → Idea Supply → Route Selection → Structure & Causality → Character State → Language Encoding → Reader Reconstruction → Load & Execution → Self-Check.
Do not jump to a later layer until the earlier ones are sufficiently stable.
1. Task Interpretation
Visible symptoms: off-topic writing, forced use of stimulus material, a polished story that never quite answers the task.
First test: Before writing, can the student explain in plain language what the topic requires, what the supplied material contributes and what remains open to invention?
If not, do not begin with vocabulary or model compositions. Repair decoding first.
Route: Visual Prompts → turn supplied evidence into usable possibilities
2. Idea Supply
Visible symptoms: “I don’t know what to write,” repeated plots, dependence on memorised model stories.
First test: Given one topic, can the student deliberately produce several relevant possibilities from memory, observation, stimulus, motives and constraints?
If not, build the warehouse of reusable mechanisms rather than handing over a finished plot.
Route: Idea Generation → Inputs → Connections → Alternatives → Routes
3. Route Selection
Visible symptoms: too many ideas, overplanning, changing plots halfway through, choosing the most dramatic idea rather than the most controllable.
First test: Can the student compare several routes by relevance, causality, motive, language load, scale and resolution—and then stop searching?
Routes: Divergence & Convergence → Expand → Compare → Reduce → Commit and Brainstorming → Open the Search → Fence the Search → Choose
4. Structure and Causality
Visible symptoms: event dumping, random twists, rushed endings, coincidence solving the main problem, a climax that does not change anything.
First test: Can the entire story be reduced to five connected states?
Situation → Pressure → Decision → Consequence → Resolution.
If one arrow is missing, repair the relationship before adding prose.
Route: Narrative Structure → build the load-bearing frame
5. Character State
Visible symptoms: characters act only because the plot needs them to, emotions appear without cause, dialogue does not match the relationship.
First test: Before the important decision, can the student state what the character wants, knows, believes, fears or misunderstands?
If the action cannot be derived from the character state and pressure, the character is being pushed by the author rather than moving through the story.
Route: Characterisation → connect motive, evidence and action
6. Language Encoding
Visible symptoms: repetitive words, forced vocabulary, awkward complex sentences, grammar failure while ideas remain good.
First test: Can the student say clearly what they mean in simpler language?
If the meaning itself is unclear, vocabulary expansion is not yet the repair. If the meaning is clear but the language cannot encode it accurately, then work on retrieval, sentence control and precision.
Routes: Vocabulary for Composition → build a live language inventory and Sentence Structure → control the encoding layer
7. Reader Reconstruction
Visible symptoms: “But I know what I meant,” missing transitions, unexplained emotional changes, excessive description that still does not make the scene clearer.
First test: Give the passage to a reader who knows nothing outside the page. What world do they reconstruct?
The gap between intended state and reconstructed state tells you what information is missing, misleading or overloaded.
Route: Imagery → Observe → Select → Compress → Evoke
8. Load and Execution
Visible symptoms: a student performs well in drills but loses grammar, structure or imagery during a full composition; quality drops sharply when time pressure is added.
First test: Remove one source of load. Does the skill return?
If sentence accuracy returns when plot planning is removed, the student may know the grammar but not yet have enough automated control to carry both tasks simultaneously.
Route: Advanced Composition → Combine → Load-Test → Refine → Automate
9. Examination Execution
Visible symptoms: good practice pieces but weak exam output, overplanning, freezing, unfinished stories, excessive late rewriting.
First test: Can the student independently move through interpretation, route selection, planning, drafting, monitoring and checking without live adult prompts?
Routes: Composition Paper 1 Runtime → execute the writing task and Exam Day Runtime → protect whole-paper execution
10. Self-Check and Repair
Visible symptoms: the same mistakes return after feedback, the student can explain corrections but does not detect them independently, every draft requires adult repair.
First test: Can the learner find one recurring error, explain why it is an error, repair it and avoid it in a changed task?
Routes: Constructive Feedback → Diagnose → Repair → Retest and Composition Repair Lab → Feedback → Rewrite → Transfer
Symptom → Earliest Test → Route
| Visible symptom | Test first | Likely route |
|---|---|---|
| Off-topic | Can the student state the actual task? | Task interpretation |
| No ideas | Can several relevant candidate routes be generated? | Idea supply |
| Too many ideas | Can weak routes be rejected quickly? | Convergence |
| Random plot | Can five causal states be connected? | Narrative structure |
| Flat characters | Is motive/state known before action? | Characterisation |
| Forced vocabulary | Is the intended meaning clear first? | Language encoding |
| Purple prose | Does each detail change reader reconstruction? | Imagery selection |
| Weak ending | Was a consequence created earlier? | Causal spine |
| Errors only under time | Does accuracy return when load is reduced? | Automation/load |
| Good only with tutor | Can the student self-route without prompts? | Independence |
One Repair at a Time
A composition can expose ten weaknesses at once. Giving ten instructions at once does not necessarily produce ten repairs. It can increase the learner’s coordination load until none of the advice survives into the next piece.
A stronger loop is:
Find earliest weak link → Select one repair → Practice it in reduced load → Reintegrate → Retest → Move only when stable.
This is diagnostic teaching rather than advice accumulation.
Three Failure Zones
| Zone | What is failing | Typical response |
|---|---|---|
| Foundation | A component is not yet installed. | Teach and practise the component in isolation. |
| Integration | The component exists but collapses when combined with others. | Reduce load, reconnect and raise complexity gradually. |
| Execution | The integrated system exists but fails under independence or examination conditions. | Rehearse routing, timing, recovery and handover. |
These zones prevent a common category error: giving foundation lessons to an execution problem, or giving exam tricks to a learner whose underlying writing system is not yet installed.
The Parent and Tutor Bandwidth Rule
Parents, teachers and tutors can all be correct and still overload the learner if they issue different repairs at the same time.
Before adding another instruction, ask:
- What is the current highest-priority weak link?
- Who is responsible for repairing it?
- What other advice can temporarily wait?
- What evidence will tell us the repair has worked?
The purpose of support is to reduce uncertainty, not create another traffic jam around the learner.
The Handover Test
A repair is not complete merely because the student can perform it while the tutor is pointing at the problem.
- Can the learner recognise the problem?
- Can they choose the appropriate repair?
- Can they execute it without prompting?
- Can they transfer it to a different topic?
- Does it survive a realistic writing load?
When those conditions hold, the scaffold can start to disappear.
Control Tower Decision
Do not ask, “What else can we teach this student?” until you have asked, “What exactly is preventing the current system from producing the intended writing?”
Then route the smallest justified repair.
Diagnose → Reduce → Repair → Reintegrate → Retest → Handover.
Need the Current Examination Specification?
This page diagnoses the writing machine. Questions about the current examination format, syllabus and assessment requirements belong in the specification layer rather than being mixed into the diagnostic runtime.
Continue to the PSLE English Examination Format → verify what the current examination contains