Grammar Evidence Card: From Rule Recognition to Real Expression
Track one high-impact structure per card. Copy it into a private note and keep the original sentence plus revision reason. Do not record sensitive customer, company, school, medical, or third-party material; use synthetic facts or a minimised excerpt.
1. Define the Real Task
markdown
Date:
Situation: email / explanation / dialogue / retelling / interview / other
Audience:
What the task must let the other person understand or do:
Time, tool, and length restrictions:
Raw sample location:2. Locate One Structural Problem
markdown
My original sentence/segment:
Most likely interpretation:
Meaning I intended:
Impact: meaning-changing / task-blocking / recurring / register mismatch / low-impact
One structure tracked in this cycle:| Sample | Structural choice | Real relationship expressed | Fits the task? |
|---|---|---|---|
| Original | |||
| Contrast A | |||
| Contrast B |
3. Write a Meaning Contrast Card
markdown
Meaning I need to express:
Forms I confuse:
Example and source from reliable material:
One-sentence explanation in my words:
Counterexample or boundary:The explanation should answer why the form matches the fact rather than copy only a terminology definition.
4. Preserve Revision and Reason
| Version | Condition | Sentence/output location | What changed | Why |
|---|---|---|---|---|
| First | No prompt | |||
| Revised | Explanation/feedback seen | |||
| Independent repeat | Material closed |
markdown
How the other person understood the revision:
Problem still affecting the task:
Suggested style change I rejected:5. Retrieve across Four Skills
| Skill | Small task | Result | Evidence |
|---|---|---|---|
| Listening | Hear the relationship carried by the structure | ||
| Reading | Mark scope, reference, or logical connection | ||
| Speaking | Use under time pressure and repair after error | ||
| Writing | Compare how two versions change reader understanding |
You do not need all four every cycle. Keep at least one input observation and one active output.
6. AI and Feedback Record
markdown
Tool/reviewer and date:
Problem observed:
Explanation given:
Did it separate error, ambiguity, register, and style?
Dictionary, corpus, teacher, or reader check:
What I accepted/rejected, and why:7. Delayed Retest
| Time point | Prompt condition | New task | Independent choice | Result |
|---|---|---|---|---|
| Day 1 | Original task, no prompt | |||
| Days 3–7 | Rule and old answer removed | Parallel topic | ||
| Day 14 | New audience or time pressure | Transfer task |
markdown
Judgment retained on day seven:
Error that returned:
Where transfer failed:8. Score and Decide
Score 0–2: 0=not completed, 1=prompt-dependent or unstable, 2=completed under the current condition.
| Item | Day 1 | Day 7 | Day 14 | Evidence reason |
|---|---|---|---|---|
| Explain form–meaning difference | ||||
| Recognise it in a new sentence | ||||
| Retrieve it in output | ||||
| Self-repair after error | ||||
| Stop repeatedly changing key meaning |
markdown
Current evidence supports: continue same structure / add variation / move to another problem / pause
Evidence:
Conclusion I still cannot draw:
Smallest next task:
Next review date:Related chapters: Grammar: Let Structure Serve Meaning | Learning Principles | Evidence Chain Template