A student receives a marked comprehension paper with three red crosses and writes, “Need to be more careful.” Another receives an essay covered in corrections and decides to “improve grammar”. Both statements sound responsible. Neither identifies what should happen next.
Error analysis turns a mark into a diagnosis. The purpose is not to catalogue every imperfection. It is to find the earliest useful decision that failed, repair it with a bounded exercise and then test whether the repair survives a new task.
A correction becomes learning only when the student can reproduce the right decision without the correction being present.
What this guide develops
This guide completes the current Secondary 2 English batch within the Secondary English Learning Hub. It works with Study-to-Exam Transfer: Planning, Timing and Self-Checking and connects back to the specific reading, writing, vocabulary, grammar, listening and oral guides whenever the diagnosis identifies a capability gap.
The marked-work examples below are fictional eduKate teaching cases. The symbols, scores and teacher comments are not official marking conventions. In real school work, use the teacher’s actual feedback and task criteria as the source of evidence.
Marks are outputs, not diagnoses
A mark tells you what happened on one assessment. It does not automatically tell you why. Two students with the same score may need different repairs.
| Visible result | Possible underlying cause |
|---|---|
| Wrong comprehension answer | Misread command, weak vocabulary, wrong reference, unsupported inference or incomplete answer. |
| Short writing | Weak planning, limited examples, slow sentence production or time loss. |
| Long vague writing | Poor paragraph function, weak evidence or repetition. |
| Grammar errors | Agreement, tense, sentence boundary, reference or rushed checking. |
| Weak oral response | Idea generation, retrieval speed, listening to follow-up or organisation. |
| Unfinished paper | Timing, one difficult item, slow reading, perfectionism or late planning. |
The first repair should target the earliest useful cause supported by evidence, not the broadest label available.
The error-analysis chain
- Observe: What exactly is wrong?
- Locate: Where did the answer first leave the correct path?
- Classify: What kind of decision failed?
- Repair: What is the smallest exercise that targets that decision?
- Return: Can the student do it again after a delay?
- Transfer: Can the student use it on unfamiliar material?
This sequence keeps correction from becoming punishment through volume. More worksheets are not automatically the answer.
Original marked-work case A: The half comparison
Question: How does Mira’s attitude to revising the display change between the beginning and end of the passage?
Student answer: “She is willing to admit the mistake and show the correction.”
Teacher comment: “Later attitude explained. What was her earlier attitude?”
The visible answer is partly correct. The first weak link is not vocabulary or inference. It is task coverage: a change question requires both sides. The repair should therefore target comparison structure.
Bounded repair: Give three pairs of short sentences and ask the student to state “initially… later…” without writing full comprehension answers. Then use one unfamiliar passage and remove the prompt.
Writing five more full passages before repairing the comparison decision would add volume around the wrong problem.
Original marked-work case B: The unsupported character label
Question: What does Daniel’s action suggest about his immediate response?
Student answer: “Daniel is a dishonest person who always hides problems.”
Teacher comment: “Too broad. Explain this moment only.”
The error is scope. The student has converted one action into a permanent character judgement and added “always” without evidence.
Bounded repair: Contrast pairs such as “tempted to hide this mistake” versus “always hides mistakes”, “hesitant in this conversation” versus “an indecisive person”. Ask which statement the evidence supports and why.
The delayed return should use a different character and a different action. If the learner still avoids permanent labels without prompting, the scope repair is transferring.
Original marked-work case C: Grammar or meaning?
Student sentence: “Because the activity is free, places are limited.”
The grammar is acceptable. The logic may be wrong if the source only states that the activity is free and separately states a participant limit. The student has created causation.
Diagnosis: Connector meaning, not sentence boundary. The repair should compare because, although, and and therefore using the same pair of facts.
This case shows why “grammar error” is sometimes too broad. The sentence can be well formed and still misrepresent the relationship.
Original marked-work case D: The essay that repeats itself
Paragraph 1 claim: Digital messages are convenient because they are easy to access.
Paragraph 2 claim: Digital messages are useful because they are convenient.
Paragraph 3 claim: Technology makes communication easier and more convenient.
The writing may be long, but the reasoning has one idea. The diagnosis is paragraph differentiation and development, not word count.
Bounded repair: Before rewriting the essay, require three distinct paragraph jobs: updating information, access conditions and physical-location use. For each, write one claim and one example only. Then rebuild the response.
Use Discursive Writing: Claims, Counterarguments and Judgement if this pattern persists.
Original marked-work case E: The unfinished paper
A student leaves the final two comprehension questions blank. The teacher notices that Question 5 contains four crossed-out answers.
The visible error is incomplete work. The evidence suggests a timing and recovery problem centred on one difficult item. The next test should not begin by increasing reading speed.
Bounded repair: Give six short questions and a stop-and-return rule. The learner must leave a usable attempt, mark uncertainty and move on after a set practice threshold. Then measure whether later questions are protected.
If the student still runs out of time even without one expensive question, investigate reading speed or writing rate next. Diagnosis should remain evidence-led.
Error categories that are useful enough to act on
| Category | Example record | Repair direction |
|---|---|---|
| Task | Answered only one side of comparison. | Question-job practice. |
| Evidence | Used true detail that did not support the claim. | Claim–evidence matching. |
| Scope | Changed “some” into “all”. | Qualifier preservation. |
| Reference | Pronoun had two possible referents. | Reference tracking. |
| Logic | Sequence changed into cause. | Connector comparison. |
| Development | Three paragraphs repeated same reason. | Paragraph-function planning. |
| Language | Subject–verb agreement failed across long phrase. | Subject-head identification. |
| Timing | One item consumed excessive time. | Checkpoint and recovery practice. |
| Retrieval | Knew vocabulary in review but could not use it orally. | Delayed active recall. |
Avoid labels that hide the decision
“Careless”, “weak English”, “bad grammar” and “poor comprehension” are too broad to guide the next action. Replace them with observable decisions.
Instead of “careless”, write “copied the old venue after the speaker corrected it”. Instead of “bad grammar”, write “used plural verb because nearest noun was plural, but subject head was singular”. Instead of “poor comprehension”, write “inference exceeded evidence boundary”.
Specific language makes the repair smaller and more testable.
The first-weak-link principle
Suppose a student writes a weak situational email. The final grammar is uneven, but the response also recommends an option that cannot accommodate the whole class. Which problem should be repaired first?
The capacity error is upstream because it changes the decision. Perfect grammar would only make the wrong recommendation clearer. Repair source reading and task fulfilment before polishing every sentence.
This does not mean grammar is unimportant. It means repair order matters. Fix the decision that causes the largest downstream failure first.
One error can have multiple causes
A wrong vocabulary-in-context answer might come from not knowing the word, ignoring the sentence, choosing a synonym with the wrong connotation or rushing because time is low. Do not infer the cause from the mark alone.
Ask the learner to explain the reasoning before showing the answer. The explanation reveals whether the concept was missing or the execution failed.
One cause can create multiple visible errors
A reference problem can produce wrong comprehension answers, confusing summaries and unclear writing. A weak distinction between possibility and certainty can appear in reading inference, situational writing and argument.
When the same underlying decision appears across tasks, repair the principle and then test it in each form. Do not treat the errors as unrelated simply because they occur in different sections of English.
The bounded repair
A bounded repair is small enough that the target skill is visible. If the problem is distinguishing two reasons from two paraphrases of one reason, use four short examples before returning to a long passage. If the problem is subject–verb agreement, do not immediately assign a full essay.
The repair should be difficult enough to require the decision but simple enough that unrelated demands do not hide it.
The return test
Immediately after correction, give one new item. Then return after several days. The second return is more informative because the answer is no longer active in short-term memory.
Record whether the learner identified the skill independently. If the tutor had to say “Remember the two sides of comparison”, the final answer may be correct but the method is still prompt-dependent.
The transfer test
Change the topic, vocabulary or text type while preserving the underlying decision. If the repair concerns qualifiers, move from “some/all” in a reading passage to “may/must” in a notice. If the repair concerns evidence scope, move from a student survey to a campaign poster.
A repair is stronger when it survives surface change. Familiarity with one corrected sentence is not the endpoint.
Build an error ledger that stays small
Do not keep hundreds of corrections. Keep recurring, high-value decisions. A useful record has four fields: original error, diagnosis, repaired example and return date.
| Error | Diagnosis | Repair | Return |
|---|---|---|---|
| Only later side of comparison | Task coverage | Initial → later pair | New passage Friday |
| May → will | Certainty drift | Modal contrast | New notice Monday |
| List of names were | Subject head | List was / names were | New sentence Wednesday |
Archive a pattern once it remains stable across several transfers. The ledger should guide attention, not become another burden.
Use confidence as data, not proof
Ask the student to mark answers as high, medium or low confidence before checking. A confident wrong answer deserves attention because the learner’s internal check failed. A low-confidence correct answer may need consolidation because the method is not yet trusted.
The goal is calibration: confidence should gradually track actual reliability. Do not punish honest uncertainty.
Original mixed diagnostic set
- 1. A student answers “everyone” when the passage says “many”.
- 2. A student writes three examples but no explanation of how they support the claim.
- 3. A student knows the correct oral vocabulary in notes but cannot retrieve it during speaking.
- 4. A student changes “although” to “because” while combining sentences.
- 5. A student spends fifteen minutes on one low-value question and leaves two later questions blank.
Possible diagnoses: 1. scope; 2. claim–evidence connection; 3. retrieval; 4. logic preservation; 5. timing and recovery. Each should lead to a different repair.
When the same error does not repeat
Do not overreact to one isolated mistake. If a student normally controls agreement but makes one slip in a rushed final paragraph, note it and look for recurrence. A new programme may be unnecessary.
Repeated errors across different tasks provide stronger evidence of a stable weak link. Frequency, context and cost all matter.
When a repair creates a new problem
A student taught to “use more evidence” may begin copying long quotations without explanation. A student taught to “write longer paragraphs” may repeat the same point. Every repair should therefore include a boundary.
For evidence: choose the smallest detail that supports the claim and explain the relationship. For development: add a new reason, example, consequence or qualification, not more words with the same function.
Independent repair challenge
Choose one recent English error from real marked work. Write the exact teacher comment or correction. Then answer four questions: What was the visible error? What decision caused it? What is the smallest repair? What unfamiliar task will test the repair later?
Do the bounded repair. Wait at least a day before the transfer test if possible. Record whether you succeeded without seeing the original correction.
A two-week error-analysis cycle
Week one: select no more than two high-value recurring errors. Repair each with short targeted practice. End the week with one unfamiliar task that contains both skills.
Week two: remove prompts and increase load slightly. Add timing only after the skill is reliable without it. At the end, decide whether the error remains active, is improving or can be archived.
Do not keep adding new repair targets faster than the learner can stabilise them. A short active list protects attention.
For parents and teachers
Ask the student what they think went wrong before explaining. Compare that self-diagnosis with the evidence. If the learner says “careless”, ask for the exact decision that should have been different.
Do not rewrite the whole composition to show a perfect version when one or two changes will expose the core lesson. Preserve the student’s ownership of the text. Correction should make the learner more independent, not more dependent on an adult editor.
Track the repair, not just the next mark. A similar score can hide improved timing, stronger task coverage and a new vocabulary problem. Look at what changed inside performance.
Connect diagnosis to the right guide
If the error is question control, return to Comprehension Question Control and Answer Precision. If the issue is meaning preservation, use Synthesis and Transformation. If the issue is evidence judgement, use Non-Fiction Reading. If the skill works but collapses under time or unfamiliarity, use Study-to-Exam Transfer.
Return to the Secondary English Learning Hub for the complete route.
Observe the error. Find the first weak link. Repair only what needs repair. Return after a delay. Then change the surface and see whether the student can carry the method alone.
