Composition feedback and self-assessment should help a writer make a better next decision. A mark can summarise performance, but improvement requires diagnosis: where did the reader’s model break, which writing layer caused the failure, what repair would change the draft, and what practice would prevent the same problem from returning?
Learners searching for a composition writing rubric, composition feedback, self-assessment checklist, how to mark a composition and how to improve composition scores often receive long lists of criteria. Lists can help, but a useful assessment system must prioritise. If the plot is irrelevant to the prompt, polishing adjectives is downstream work. If the story architecture is sound, sentence and language feedback can become more specific.
This world-facing guide owns the general composition feedback, rubric and self-assessment mechanism while preserving eduKateSG’s existing PSLE-specific marking, feedback and success-criteria pages. Exact scoring schemes vary by school and examination authority; use the official rubric for formal assessment. The improvement engine here is OBSERVE → LOCATE → CLASSIFY → PRIORITISE → REPAIR → RETEST → PRACTISE → TRANSFER.
Quick Read: A Useful Composition Rubric
Judge in layers: relevance and meaning, story architecture and development, character and viewpoint, paragraph and sentence clarity, then vocabulary, grammar, punctuation and spelling. Do not average away a load-bearing failure. Find the earliest weak link and repair that first.
1. Feedback Is Information for the Next Decision
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
2. Separate Assessment from Improvement
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
3. Read the Whole Composition First
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
4. Check Prompt Ownership
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
5. Check Causal Architecture
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
6. Check Character Agency
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
7. Check Development
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
8. Check Climax Ownership
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
9. Check Consequence
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
10. Check Point of View
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
11. Check Setting Function
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
12. Check Paragraph Coherence
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
13. Check Sentence Clarity
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
14. Check Vocabulary Control
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
15. Check Grammar Control
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
16. Find the Earliest Weak Link
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
17. Prioritise One Structural Target
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
18. Prioritise One Language Target
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
19. Use Success Criteria Before Writing
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
20. Use Success Criteria After Writing
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
21. Use Self-Assessment Before Teacher Feedback
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
22. Use Teacher Feedback as Calibration
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
23. Use Peer Feedback with Narrow Questions
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
24. Use Evidence, Not Praise Labels
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
25. Turn Comments into Actions
Feedback becomes useful when it names evidence. “Good story” is pleasant but hard to reproduce. “The torn-tape detail is planted before the climax and later identifies the parcel” tells the learner what mechanism succeeded. The same standard applies to corrective feedback.
Separate observation from judgment. First state what the text does: the protagonist searches three times but each attempt leaves the options unchanged. Then classify: flat escalation. Then prescribe: make the second attempt reveal information or increase cost. This sequence makes the feedback auditable.
Adrian, Jo, Ben, Aisha, Ryan, Mira, Clara and Ethan should not receive identical feedback packages. A learner with strong architecture and weak sentence boundaries needs a different next task from a learner with fluent sentences and irrelevant plots. Routing is more useful than generic completeness.
Run the weak-signal test. Twenty comments on one page predict poor prioritisation. A score without evidence predicts low transfer. Rewriting the learner’s sentences predicts dependency. Repeated praise for vocabulary while endings remain unfinished predicts compensation across layers. Feedback itself can fail.
Run the non-compensatory gate. Do not let a strong vocabulary score hide a missing climax, or a creative premise hide unreadable sentence boundaries. Some failures affect the whole reader experience. Diagnose gates before averaging surface strengths.
Use the next-draft test. After reading the feedback, can the learner state what to do differently before writing the next composition? If not, the comment may describe the past without improving the future. Every important comment should route to a prevention cue or practice task.
Use learner explanation. Ask the writer to paraphrase the feedback and propose a repair. Misunderstanding becomes visible immediately. If the teacher supplies every correction, the page improves faster than the writer. Calibration requires learner participation.
Finally, retest under changed content. The repair is not complete because one corrected paragraph looks better. A fresh prompt reveals whether the learner can retrieve the principle independently. Transfer is the real receipt for feedback.
26. Avoid Marking Everything
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
27. Avoid Vague Praise
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
28. Avoid Vague Criticism
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
29. Avoid Rewriting the Student’s Work
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
30. Avoid Rubric Overload
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
31. Avoid Score-Chasing Without Diagnosis
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
32. Avoid Comparing Writers Instead of Drafts
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
33. Avoid Rewarding Length Automatically
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
34. Avoid Rewarding Rare Vocabulary Automatically
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
35. Avoid Penalising Controlled Simplicity
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
36. Build a Relevance Rubric
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
37. Build a Causality Rubric
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
38. Build a Character Rubric
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
39. Build a Setting Rubric
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
40. Build a Viewpoint Rubric
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
41. Build a Pacing Rubric
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
42. Build a Paragraph Rubric
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
43. Build a Sentence Rubric
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
44. Build a Vocabulary Rubric
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
45. Build a Grammar Rubric
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
46. Build a Revision Rubric
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
47. Build a Transfer Rubric
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
48. Use Exemplars for Calibration
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
49. Use Counterexamples for Calibration
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
50. Use Anchor Descriptions
A rubric dimension is useful when it corresponds to something the reader can observe. Replace broad labels with evidence: “events form a causal chain and the protagonist’s choice changes the outcome” is more diagnostic than “plot: excellent”. Descriptors should tell the learner what functioning looks like.
Keep dimensions distinct enough to route repair. If “content” includes relevance, character, plot, theme and creativity in one box, a low result gives little guidance. Separate load-bearing systems conceptually even if an official assessment later combines them into broader scores.
Use levels as descriptions, not identities. A draft can show strong paragraph coherence and weak causal escalation. Do not convert a performance state into “weak writer”. The unit of diagnosis is the text and the current skill under current conditions.
Calibrate with examples. Show two passages that differ on one dimension and ask what evidence changes the judgment. Agreement improves when raters share observable anchors. This is useful for teachers, parents and learners using self-assessment.
Avoid false precision. A locally invented numerical score may look objective while hiding uncertain judgment. Where no official scoring scheme applies, descriptive categories and evidence can be more honest. For formal assessment, use the relevant authority’s actual rubric rather than inventing a replacement.
Use comments to route practice. If the dimension is viewpoint, the next exercise may be a knowledge ledger. If it is climax agency, practise protagonist-owned decisions. A rubric should connect assessment to curriculum rather than end at classification.
Check interdependence. A weak setting may create a weak conflict; ambiguous pronouns may make character tracking appear weak. Diagnose the earliest cause rather than treating every downstream symptom as an independent deficit.
Finally, revisit the same dimension after new writing. The purpose of a rubric is not merely to describe one composition. It should help reveal whether the underlying capability is becoming more reliable across prompts.
51. Full Feedback Workshop: Diagnose Before Correcting
Draft situation: Mira mislabels a parcel, searches three stacks, a stranger finds the correct box, everyone praises her and she writes that honesty is important. The sentences are mostly grammatical. A surface-first marker might correct vocabulary and punctuation. A diagnostic marker should first ask whether the story’s architecture works.
Observation: Mira does not make the decisive action that solves the problem. Classification: protagonist agency is weak. Effect: the climax feels coincidental. Priority: high, because it affects the central story promise. Repair: make Mira’s admission unlock shared information that leads to the box. Practice: write three protagonist-owned climaxes.
Observation: the three search attempts do the same thing. Classification: flat escalation. Effect: the middle feels repetitive. Repair: let attempt one remove the easy solution, attempt two produce poor information because Mira conceals the reason, and the van’s arrival close time. Practice: escalation ladders where every beat changes options.
Observation: “Honesty is important” appears only in the final moral. Classification: theme is pasted on. Repair: make concealment itself worsen the information problem, then show changed behaviour in the ending. Theme becomes causal evidence rather than commentary.
Observation: two sentences contain vague pronouns. Classification: reference clarity. Priority: medium, after architecture. Repair: repeat the noun or restructure. This demonstrates prioritisation: grammar still matters, but it is not the earliest weak link.
52. A Layered Composition Rubric
Layer 1 — Prompt and meaning: Does the composition answer the task and make the topic operational? Layer 2 — Architecture: Is there a causal chain, escalation, decisive turn and consequence? Layer 3 — Character and viewpoint: Are goals, information and choices coherent? Layer 4 — Organisation: Do paragraphs and pacing guide attention? Layer 5 — Language: Are vocabulary and sentences precise and readable? Layer 6 — Accuracy: Are grammar, punctuation and spelling controlled?
53. Four Descriptive Performance States
Not yet visible: the target mechanism is missing or cannot be judged. Emerging: the mechanism appears but fails under pressure or needs substantial support. Functional: it works reliably in the current draft with minor weaknesses. Transfer-ready: it works independently across changed prompts and the learner can diagnose or reproduce it. These are learning states, not official examination grades.
54. Self-Assessment Before Submission
Before asking for external feedback, the learner should answer: What is my story promise? Where is the climax? What choice belongs to the protagonist? What changes afterwards? Which paragraph is weakest? Which sentence is hardest to parse? What is one error I know I often make? This creates an internal prediction that external feedback can calibrate.
55. Self-Assessment After Feedback
Compare prediction with the teacher’s diagnosis. Did both locate the same weakness? If not, why? A learner who thought vocabulary was the problem may discover that the real failure was relevance. Calibration is the gradual alignment between self-judgment and informed external judgment.
56. Feedback That Preserves Agency
Prefer questions that force reconstruction: “What does Mira know here?” “What changed after this attempt?” “Who makes the decisive choice?” “What would happen if this paragraph disappeared?” These questions narrow the search without supplying the answer. The learner remains the writer.
57. Feedback on Ideas
Do not replace the learner’s premise simply because another story would be easier to write. First test whether the existing idea can satisfy relevance, causality and consequence. Preserve creative ownership where possible. Suggest a new premise only when the current one cannot reasonably carry the task or the learner explicitly wants alternatives.
58. Feedback on Vocabulary
Mark words for meaning, collocation, register and precision—not for failing to sound advanced. Praise a simple exact word when it performs the job. If an ambitious word is misused, ask the learner to paraphrase it and compare alternatives. Vocabulary feedback should increase control, not lexical anxiety.
59. Feedback on Grammar
Do not mark every repeated instance indefinitely. Correct or identify the first few, classify the pattern and ask the learner to find the remaining examples. This shifts detection responsibility gradually. Record recurring systems in a personal error corpus for later retrieval practice.
60. Feedback on Creativity
Creativity is difficult to reduce to novelty alone. Look for useful combination, specificity, causal use of setting, fresh attention, credible choices and transformation of familiar material. A quiet but well-observed story may be more creatively controlled than a spectacular plot assembled from random surprises.
61. Feedback on Voice
Voice should not become a demand for decorative individuality. Look for stable attention, sentence tendencies, vocabulary fit and relationship to the narrator’s perspective. Preserve authentic clarity. Do not rewrite a student’s controlled sentence into an adult’s preferred style and call that voice development.
62. Feedback Conference Protocol
First ask the writer what they were trying to achieve. Second ask where they think the draft is weakest. Third present one piece of evidence from the text. Fourth agree on one high-impact repair. Fifth let the writer attempt it. Sixth identify one practice task for transfer. A short conference can be powerful when it stays diagnostic.
63. Parent Support Protocol
Parents can help by asking for the story’s goal, climax and ending rather than supplying sentences. If a child is stuck, ask one discriminating question at a time. Praise observable decisions—clearer consequence, stronger setup, precise verb—rather than global labels such as “talented”. Support should make the learner more independent at the next sitting.
64. Teacher Marking Protocol
Read once without marking to understand the whole composition. On the second pass, identify the earliest high-impact weakness. Mark a limited number of representative language errors. Add one success evidence and one next action. Where formal scoring is required, score according to the official rubric, but keep instructional feedback focused enough to be usable.
65. Peer Feedback Protocol
Peers need narrow questions. “Where were you confused?” “Which action changed the story most?” “What detail did you expect to matter later?” are safer than asking classmates to grade the entire composition. Peer feedback is useful for reader response; technical correction should be calibrated carefully.
66. The Feedback-to-Practice Router
Irrelevance → prompt decoding drills. Flat plot → causal five-beat plans. Weak agency → climax ownership drills. Rushed ending → consequence writing. Head-hopping → knowledge ledgers. Generic description → viewpoint selection. Choppy sentences → combining. Overloaded sentences → splitting. Tense drift → timeline practice. Vocabulary misuse → meaning and collocation retrieval.
67. The Feedback-to-Next-Draft Cue
Convert each important comment into a pre-writing cue. “Ending rushed” becomes “write ending beat before drafting”. “Climax coincidental” becomes “plant solution evidence in setup”. “Pronouns unclear” becomes “check every he, she, it, they after drafting”. Prevention is feedback moved upstream.
68. A Student-Friendly Self-Assessment Card
Before I finish: Does my story answer the prompt? Can I explain what my character wants? Does each problem change something? Does my character make the important choice? Did I show what happened because of it? Can the reader tell who is speaking and acting? Is any sentence confusing? Did I check my usual grammar mistakes?
69. A Teacher Diagnostic Card
Locate: prompt, goal, obstacle, escalation, climax, consequence. Measure: which link is weakest? Narrow: choose one structural and one language target. Explain: show evidence from the text. Act: learner repairs. Observe: compare before and after. Transfer: assign a changed-content retest.
70. A Parent Diagnostic Card
Ask: What is this story really about? Where does the problem begin? What choice is hardest? What changes at the end? Which sentence are you least sure about? Let the child answer before suggesting anything. The aim is to activate self-explanation, not to become the co-author.
71. Calibration with Model Compositions
Use models to clarify criteria, not to demand resemblance. Compare how two strong compositions can solve the same prompt differently. Identify common mechanisms—relevance, causality, agency, consequence—while preserving different voice, setting and plot. Calibration should widen the learner’s design space, not narrow it to one template.
72. Calibration with Weak Examples
Deliberately weak examples are valuable because failure is easier to see. Remove setup from a climax, introduce a random rescuer or attach a generic moral. Ask learners to diagnose the earliest failure and repair it. Error contrast builds rubric understanding faster than praise alone.
73. Feedback and Motivation
Feedback should make improvement legible. Show what already works, but tie praise to evidence. Limit corrective load to what the learner can act on. A page covered in marks may communicate effort from the teacher while giving the learner no clear route. Precision is kinder and more effective than volume.
74. Feedback and High Performers
Advanced writers still need diagnosis. Their bottlenecks may involve over-complex plots, subtle viewpoint leakage, overwritten description or weak thematic restraint. Do not respond to fluent prose only with praise. Find the next discriminating constraint while preserving strengths.
75. Feedback and Struggling Writers
Reduce the number of simultaneous targets. Secure a complete causal story before demanding stylistic range. Use short models, completion tasks and immediate feedback. Celebrate specific control—one clear paragraph, one correct tense sequence—then expand. Reliability grows from stable foundations.
76. Feedback and Timed Writing
A timed draft contains process evidence. Note where the writer ran out of time, where the ending compressed or where sentence errors increased. The repair may be time allocation rather than knowledge. Compare untimed and timed performance to separate capability from execution.
77. Feedback and Revision
Require the learner to submit or preserve both versions when possible. The difference between drafts is evidence of learning. Ask which change had the greatest downstream effect. Revision should become an object of assessment, not an invisible cleanup stage.
78. Feedback and Transfer
After a successful repair, change the content. If the learner can reproduce the mechanism on a fresh prompt without cues, the feedback has transferred. If not, restore a smaller scaffold and repeat. The corrected original is not sufficient evidence of mastery.
79. Useful eduKateSG Routes
Continue with Composition Practice System, Prompt to Final Draft, Revision and Proofreading, Diagnostic Assessment, Self-Evaluation, PSLE Self-Assessment and Success Criteria and Feedback in PSLE Composition Writing.
80. Final Principle
The purpose of composition feedback is not to prove that an evaluator can find errors. It is to improve the writer’s internal evaluator. Locate the earliest weak link, show the evidence, make the learner repair it, retest under changed conditions and move the successful correction upstream into the next draft.
81. Relevance Calibration Laboratory
Prepare three short passages that differ mainly on this dimension: one where the mechanism is missing, one where it is emerging and one where it functions reliably. Remove names or labels that reveal the intended ordering. Ask the learner to rank the evidence, then explain the judgment.
Next reveal the anchor descriptions and compare reasoning. Disagreement is useful. Identify whether the learner noticed a different feature, misunderstood the criterion or weighted a downstream strength too heavily. Calibration improves through explicit comparison of evidence.
Now repair the weakest passage. Do not copy the strongest one. Create a different solution that satisfies the same criterion. This proves the rubric describes function rather than one preferred wording or plot.
Transfer to the learner’s own composition. Find one passage that sits somewhere on the same continuum. Ask the learner to locate the evidence before the teacher comments. Self-assessment becomes increasingly accurate when external judgment is used as calibration rather than replacement.
Delay the next check. Several days later, present a fresh passage without the rubric visible. Ask for classification and evidence. If judgment remains stable, the criterion is becoming internalised. If not, return to contrastive examples.
Keep official marks separate where necessary. A school or examination may combine dimensions differently or apply specific task rules. Use those official criteria for formal scoring. Use this laboratory for learning the underlying writing mechanism.
Record one conditional success rule: “A climax is functional when the protagonist’s prepared choice changes the story state,” or “Vocabulary is controlled when meaning, collocation and register fit.” Conditional rules support future independent judgment.
Finally, ask the learner to design one weak example intentionally. Creating failure requires understanding the rule well enough to violate it. Then repair the example. Generation is a demanding form of rubric comprehension.
82. Causality Calibration Laboratory
Prepare three short passages that differ mainly on this dimension: one where the mechanism is missing, one where it is emerging and one where it functions reliably. Remove names or labels that reveal the intended ordering. Ask the learner to rank the evidence, then explain the judgment.
Next reveal the anchor descriptions and compare reasoning. Disagreement is useful. Identify whether the learner noticed a different feature, misunderstood the criterion or weighted a downstream strength too heavily. Calibration improves through explicit comparison of evidence.
Now repair the weakest passage. Do not copy the strongest one. Create a different solution that satisfies the same criterion. This proves the rubric describes function rather than one preferred wording or plot.
Transfer to the learner’s own composition. Find one passage that sits somewhere on the same continuum. Ask the learner to locate the evidence before the teacher comments. Self-assessment becomes increasingly accurate when external judgment is used as calibration rather than replacement.
Delay the next check. Several days later, present a fresh passage without the rubric visible. Ask for classification and evidence. If judgment remains stable, the criterion is becoming internalised. If not, return to contrastive examples.
Keep official marks separate where necessary. A school or examination may combine dimensions differently or apply specific task rules. Use those official criteria for formal scoring. Use this laboratory for learning the underlying writing mechanism.
Record one conditional success rule: “A climax is functional when the protagonist’s prepared choice changes the story state,” or “Vocabulary is controlled when meaning, collocation and register fit.” Conditional rules support future independent judgment.
Finally, ask the learner to design one weak example intentionally. Creating failure requires understanding the rule well enough to violate it. Then repair the example. Generation is a demanding form of rubric comprehension.
83. Character Calibration Laboratory
Prepare three short passages that differ mainly on this dimension: one where the mechanism is missing, one where it is emerging and one where it functions reliably. Remove names or labels that reveal the intended ordering. Ask the learner to rank the evidence, then explain the judgment.
Next reveal the anchor descriptions and compare reasoning. Disagreement is useful. Identify whether the learner noticed a different feature, misunderstood the criterion or weighted a downstream strength too heavily. Calibration improves through explicit comparison of evidence.
Now repair the weakest passage. Do not copy the strongest one. Create a different solution that satisfies the same criterion. This proves the rubric describes function rather than one preferred wording or plot.
Transfer to the learner’s own composition. Find one passage that sits somewhere on the same continuum. Ask the learner to locate the evidence before the teacher comments. Self-assessment becomes increasingly accurate when external judgment is used as calibration rather than replacement.
Delay the next check. Several days later, present a fresh passage without the rubric visible. Ask for classification and evidence. If judgment remains stable, the criterion is becoming internalised. If not, return to contrastive examples.
Keep official marks separate where necessary. A school or examination may combine dimensions differently or apply specific task rules. Use those official criteria for formal scoring. Use this laboratory for learning the underlying writing mechanism.
Record one conditional success rule: “A climax is functional when the protagonist’s prepared choice changes the story state,” or “Vocabulary is controlled when meaning, collocation and register fit.” Conditional rules support future independent judgment.
Finally, ask the learner to design one weak example intentionally. Creating failure requires understanding the rule well enough to violate it. Then repair the example. Generation is a demanding form of rubric comprehension.
84. Pacing Calibration Laboratory
Prepare three short passages that differ mainly on this dimension: one where the mechanism is missing, one where it is emerging and one where it functions reliably. Remove names or labels that reveal the intended ordering. Ask the learner to rank the evidence, then explain the judgment.
Next reveal the anchor descriptions and compare reasoning. Disagreement is useful. Identify whether the learner noticed a different feature, misunderstood the criterion or weighted a downstream strength too heavily. Calibration improves through explicit comparison of evidence.
Now repair the weakest passage. Do not copy the strongest one. Create a different solution that satisfies the same criterion. This proves the rubric describes function rather than one preferred wording or plot.
Transfer to the learner’s own composition. Find one passage that sits somewhere on the same continuum. Ask the learner to locate the evidence before the teacher comments. Self-assessment becomes increasingly accurate when external judgment is used as calibration rather than replacement.
Delay the next check. Several days later, present a fresh passage without the rubric visible. Ask for classification and evidence. If judgment remains stable, the criterion is becoming internalised. If not, return to contrastive examples.
Keep official marks separate where necessary. A school or examination may combine dimensions differently or apply specific task rules. Use those official criteria for formal scoring. Use this laboratory for learning the underlying writing mechanism.
Record one conditional success rule: “A climax is functional when the protagonist’s prepared choice changes the story state,” or “Vocabulary is controlled when meaning, collocation and register fit.” Conditional rules support future independent judgment.
Finally, ask the learner to design one weak example intentionally. Creating failure requires understanding the rule well enough to violate it. Then repair the example. Generation is a demanding form of rubric comprehension.
85. Paragraph Calibration Laboratory
Prepare three short passages that differ mainly on this dimension: one where the mechanism is missing, one where it is emerging and one where it functions reliably. Remove names or labels that reveal the intended ordering. Ask the learner to rank the evidence, then explain the judgment.
Next reveal the anchor descriptions and compare reasoning. Disagreement is useful. Identify whether the learner noticed a different feature, misunderstood the criterion or weighted a downstream strength too heavily. Calibration improves through explicit comparison of evidence.
Now repair the weakest passage. Do not copy the strongest one. Create a different solution that satisfies the same criterion. This proves the rubric describes function rather than one preferred wording or plot.
Transfer to the learner’s own composition. Find one passage that sits somewhere on the same continuum. Ask the learner to locate the evidence before the teacher comments. Self-assessment becomes increasingly accurate when external judgment is used as calibration rather than replacement.
Delay the next check. Several days later, present a fresh passage without the rubric visible. Ask for classification and evidence. If judgment remains stable, the criterion is becoming internalised. If not, return to contrastive examples.
Keep official marks separate where necessary. A school or examination may combine dimensions differently or apply specific task rules. Use those official criteria for formal scoring. Use this laboratory for learning the underlying writing mechanism.
Record one conditional success rule: “A climax is functional when the protagonist’s prepared choice changes the story state,” or “Vocabulary is controlled when meaning, collocation and register fit.” Conditional rules support future independent judgment.
Finally, ask the learner to design one weak example intentionally. Creating failure requires understanding the rule well enough to violate it. Then repair the example. Generation is a demanding form of rubric comprehension.
86. Sentence Calibration Laboratory
Prepare three short passages that differ mainly on this dimension: one where the mechanism is missing, one where it is emerging and one where it functions reliably. Remove names or labels that reveal the intended ordering. Ask the learner to rank the evidence, then explain the judgment.
Next reveal the anchor descriptions and compare reasoning. Disagreement is useful. Identify whether the learner noticed a different feature, misunderstood the criterion or weighted a downstream strength too heavily. Calibration improves through explicit comparison of evidence.
Now repair the weakest passage. Do not copy the strongest one. Create a different solution that satisfies the same criterion. This proves the rubric describes function rather than one preferred wording or plot.
Transfer to the learner’s own composition. Find one passage that sits somewhere on the same continuum. Ask the learner to locate the evidence before the teacher comments. Self-assessment becomes increasingly accurate when external judgment is used as calibration rather than replacement.
Delay the next check. Several days later, present a fresh passage without the rubric visible. Ask for classification and evidence. If judgment remains stable, the criterion is becoming internalised. If not, return to contrastive examples.
Keep official marks separate where necessary. A school or examination may combine dimensions differently or apply specific task rules. Use those official criteria for formal scoring. Use this laboratory for learning the underlying writing mechanism.
Record one conditional success rule: “A climax is functional when the protagonist’s prepared choice changes the story state,” or “Vocabulary is controlled when meaning, collocation and register fit.” Conditional rules support future independent judgment.
Finally, ask the learner to design one weak example intentionally. Creating failure requires understanding the rule well enough to violate it. Then repair the example. Generation is a demanding form of rubric comprehension.
87. Vocabulary Calibration Laboratory
Prepare three short passages that differ mainly on this dimension: one where the mechanism is missing, one where it is emerging and one where it functions reliably. Remove names or labels that reveal the intended ordering. Ask the learner to rank the evidence, then explain the judgment.
Next reveal the anchor descriptions and compare reasoning. Disagreement is useful. Identify whether the learner noticed a different feature, misunderstood the criterion or weighted a downstream strength too heavily. Calibration improves through explicit comparison of evidence.
Now repair the weakest passage. Do not copy the strongest one. Create a different solution that satisfies the same criterion. This proves the rubric describes function rather than one preferred wording or plot.
Transfer to the learner’s own composition. Find one passage that sits somewhere on the same continuum. Ask the learner to locate the evidence before the teacher comments. Self-assessment becomes increasingly accurate when external judgment is used as calibration rather than replacement.
Delay the next check. Several days later, present a fresh passage without the rubric visible. Ask for classification and evidence. If judgment remains stable, the criterion is becoming internalised. If not, return to contrastive examples.
Keep official marks separate where necessary. A school or examination may combine dimensions differently or apply specific task rules. Use those official criteria for formal scoring. Use this laboratory for learning the underlying writing mechanism.
Record one conditional success rule: “A climax is functional when the protagonist’s prepared choice changes the story state,” or “Vocabulary is controlled when meaning, collocation and register fit.” Conditional rules support future independent judgment.
Finally, ask the learner to design one weak example intentionally. Creating failure requires understanding the rule well enough to violate it. Then repair the example. Generation is a demanding form of rubric comprehension.
88. Grammar Calibration Laboratory
Prepare three short passages that differ mainly on this dimension: one where the mechanism is missing, one where it is emerging and one where it functions reliably. Remove names or labels that reveal the intended ordering. Ask the learner to rank the evidence, then explain the judgment.
Next reveal the anchor descriptions and compare reasoning. Disagreement is useful. Identify whether the learner noticed a different feature, misunderstood the criterion or weighted a downstream strength too heavily. Calibration improves through explicit comparison of evidence.
Now repair the weakest passage. Do not copy the strongest one. Create a different solution that satisfies the same criterion. This proves the rubric describes function rather than one preferred wording or plot.
Transfer to the learner’s own composition. Find one passage that sits somewhere on the same continuum. Ask the learner to locate the evidence before the teacher comments. Self-assessment becomes increasingly accurate when external judgment is used as calibration rather than replacement.
Delay the next check. Several days later, present a fresh passage without the rubric visible. Ask for classification and evidence. If judgment remains stable, the criterion is becoming internalised. If not, return to contrastive examples.
Keep official marks separate where necessary. A school or examination may combine dimensions differently or apply specific task rules. Use those official criteria for formal scoring. Use this laboratory for learning the underlying writing mechanism.
Record one conditional success rule: “A climax is functional when the protagonist’s prepared choice changes the story state,” or “Vocabulary is controlled when meaning, collocation and register fit.” Conditional rules support future independent judgment.
Finally, ask the learner to design one weak example intentionally. Creating failure requires understanding the rule well enough to violate it. Then repair the example. Generation is a demanding form of rubric comprehension.
89. The Minimal Feedback Record
For each composition, store only: one strength with evidence, one earliest weak link, one representative language pattern, one repair completed by the learner, and one next-draft cue. This compact record is enough to reveal trends across time without turning writing instruction into paperwork.
90. The Progress Review
Every four to six compositions, review the records rather than a single mark. Which failures disappeared? Which recur? Which now appear only under time pressure? Which strengths have become stable? Choose the next teaching priority from the trend. Progress is a trajectory, not a snapshot.
91. The Independence Test
Give a fresh composition back without comments and ask the learner to identify the two highest-impact revisions. Compare with the teacher’s diagnosis. Increasing overlap is evidence that the learner’s internal evaluator is improving. Eventually the teacher’s role shifts from locator to calibrator.
92. Closing Principle
A mature rubric is not a cage around creative writing. It is a set of lenses for seeing whether communication works. Use it to locate failure, not to standardise imagination. The strongest endpoint is a writer who can judge their own work accurately enough to revise before anyone else has to point out the problem.
Route: continue through What is Creative Writing for the complete writing system, return to the English Learning Hub for the wider language map, or use the Deep Reading Index for the complete long-tail library.
Continue deeper into English writing: this eduKateSG guide explains one composition mechanism. For the complete English-writing system across composition, creative writing, essays, genre, audience, revision, examination performance and transfer, continue to SETC English Writing and Composition.