A paper comes home with the same comment for the third time: “Explain your answer.” The student reads it, copies a longer model answer and puts the paper away. The page has changed. Whether the student can explain the next answer is still unknown.
That small gap is the subject of this guide. To improve learning from feedback, turn a comment into a specific change you can make, demonstrate that change on your own work, and check whether you can make it again on a fresh task. Receiving advice, agreeing with it and copying a correction are different from being able to use it.
This is the next practical route in the How to Improve series, following How to Improve Productive Struggle. It is a worked-example companion to How Feedback Works in Learning, not a replacement for that explanation.
The focus here is the student’s side of the exchange: reading a teacher’s comment, deciding what it means, repairing a particular piece of work, dealing with unclear or conflicting advice, and carrying the lesson into a later attempt. The examples, classroom conversations and practice arrangements below are original illustrations, not records of actual students, official examination questions or a tested eduKate intervention. Research findings are identified separately and linked to their sources.
Read by task: English example · Mathematics example · Data example · Parents · Research and limits.
Begin with one piece of marked work
Choose a returned assignment rather than a whole stack. Put the question, your original answer and the feedback beside one another. Keep the original visible; do not erase the evidence of how you first approached the task. You are trying to understand a difference, and a difference needs two versions.
Read the task before the comment. “Explain” means something different when the question asks for a character’s motivation, an algebraic justification or a conclusion from measurements. Feedback must be interpreted in relation to what the work was meant to accomplish.
Now finish this sentence: “On my next attempt, I will change _____ because _____.” If the blank can only be filled with “be better,” “write more” or “try harder,” the comment has not yet become an action. Ask for clarification or examine an example before beginning another large piece of work.
A usable starting routine is to locate the relevant passage or line, describe the problem in ordinary language, make one purposeful revision, and explain why it is better. Keep a fresh question for a later check. This guide develops each of those decisions rather than treating correction as a single act.
Why reading the comment is only the beginning
Carless and Boud’s 2018 conceptual paper describes student feedback literacy in terms of recognising feedback’s value, judging quality, handling emotional responses and taking action. The important distinction is between information being supplied and the learner being able to use it. Their paper concerns higher education; it is not a direct trial of the school routines proposed here. It provides a useful vocabulary for examining what a feedback exchange requires of the receiver. Research source: Carless and Boud.
For a school learner, that distinction suggests a practical question: where did the exchange stop? Perhaps the comment was not read. Perhaps it was read but misunderstood. Perhaps the student understood it but lacked the knowledge to carry it out. Perhaps the correction worked beside the model answer but not later.
Those situations should not receive the same response. Another reminder to “use the feedback” will not explain an unfamiliar term. Another model answer will not necessarily show whether the student can choose the right method. A better response begins by locating the unfinished part of the exchange.
Separate receipt, interpretation, revision and later use
Consider four statements: “I received the comments”; “I can explain what they mean”; “I revised the original”; “I used the same lesson independently elsewhere.” Each statement describes a different achievement.
A student may have reached the third statement without reaching the fourth. For example, the teacher may point to every weak sentence in an essay, making revision possible. In the next essay, the learner must also detect those weaknesses. Correcting a located problem and finding the problem are different demands.
Use these distinctions when reviewing progress. Ask the student to show one place where the comment was applied, then ask how they would recognise a similar issue without the teacher’s circle. The answer might become a small checking question: “Have I connected this evidence to my claim?” or “Did I multiply every term inside the bracket?”
These stages are a practical organising device, not a scientific scale of maturity. They help adults and learners avoid treating a neat corrected page as evidence for capabilities that have not yet been checked.
Translate feedback into a decision, not a slogan
Some comments name a desirable quality without explaining the decision that creates it. “Be precise” names an outcome. “Replace ‘it’ with the object you are discussing so the reader knows what changed” describes an action. “Develop your point” names a need. “Explain how your example supports the claim rather than repeating the claim” narrows the work.
Do not invent the teacher’s intention when a comment has several possible meanings. Mark the uncertainty and ask a targeted question. “When you wrote ‘unclear,’ do you mean that my idea is wrong, or that the reader cannot follow how I expressed it?” is more useful than silently adding three sentences.
The student does not need sophisticated terminology to do this. A younger learner can point to the sentence and say, “I think this part does not tell the reader who is speaking. Is that what you mean?” The adult can confirm or redirect.
The goal is an agreed next decision. Until that agreement exists, increased effort may improve a feature the teacher was not asking the student to change.
Worked example: “Explain your inference” in English
The following extract and question are invented for practice:
Before her name was called, Maya read her speech card for the third time. When the classroom door opened, she pressed the card between both hands and looked towards the teacher.
Suppose the question asks, “What suggests that Maya may be nervous before speaking?” A student answers, “She is nervous because she is scared.” The response uses another emotion word but does not connect the inference to the extract.
A comment such as “Use evidence” should not become “copy any sentence from the passage.” The learner needs to choose details that support the particular inference. Here, repeated reading and tightly holding the card are relevant clues. The wording “may be nervous” also matters: the question asks for a supported interpretation, not certainty about everything Maya feels.
A revised answer could be: “Maya reads her speech card repeatedly and presses it between both hands. These actions suggest that she is anxious about speaking and is trying to prepare herself.” This answer connects selected details to the interpretation.
Now examine an overcorrection: “Maya is terrified because she has always failed at public speaking and believes that everyone will laugh.” That answer is longer and more dramatic, but it invents a history and expectation that the extract does not provide. The feedback did not authorise the student to create missing evidence.
The usable lesson is therefore narrower than “write more”: select a relevant textual detail, explain its connection to the inference, and avoid adding unsupported background. The revision can become longer, shorter or approximately the same length, depending on what was missing.
Check the English lesson on a different passage
Use a new invented extract:
After the teacher announced that the missing library book had been found, Arun loosened his grip on the return slip and let out a long breath.
Ask the learner what suggests that Arun may feel relieved. Do not display the revised Maya answer beside the new question. The intended check is whether the learner selects Arun’s actions and links them to relief, not whether the learner can replace one name in a memorised sentence frame.
An acceptable response might connect the loosened grip and long breath to a release of tension after the announcement. A response claiming that Arun had stolen the book would exceed the extract. A response copying the whole passage without explaining its relevance would still leave the evidence-to-inference connection incomplete.
Compare the two attempts using the same questions: Is there a relevant detail? Is its connection explained? Has the learner stayed within what the text supports? Do not award invented official marks; this is a teaching check with stated criteria.
A successful answer is evidence of this particular skill in this particular task. It does not establish that all comprehension weaknesses have been repaired. The next lesson should follow what the learner actually demonstrated.
Worked example: “Check your expansion” in Mathematics
Consider the equation 4(2x − 3) = 20. A learner writes 8x − 3 = 20, then obtains x = 23/8. The later algebra is consistent with the line the learner wrote; the first problem is the expansion. The multiplier outside the bracket applies to both terms inside it.
Useful feedback directs attention to that step. A teacher might ask, “What is four groups of 2x minus three?” or show the intermediate form 4 × 2x − 4 × 3. The correction becomes 8x − 12 = 20, so 8x = 32 and x = 4. Substitution checks the result: 4(2 × 4 − 3) = 20.
Now ask the learner to explain the difference between the original expression and 4(2x) − 3. The second expression genuinely becomes 8x − 3 because the subtraction lies outside the bracket. This contrast reveals whether the student understands the role of the bracket rather than only remembering to change a three into a twelve.
The action statement might be: “When a number multiplies a bracket, I will show what it multiplies before simplifying.” That is a temporary support for this error, not a requirement to write the same expanded working on every easy expression forever.
A correction that changes the answer to four without explaining the first line has repaired the page but leaves the original decision untested. Keep the wrong line available for comparison and ask the learner to identify exactly what makes it invalid.
Give the Mathematics correction somewhere new to operate
A fresh check could use 3(2x + 5) = 27. The correct expansion is 6x + 15 = 27, giving x = 2. Ask for an explanation of the expansion before checking the final value. The student should not need to be told that this is another “multiply both terms” question.
A contrast item, 3(2x) + 5 = 27, gives a different equation and solution. Placing the expressions together is useful for discussing structure, but later mix them among other familiar questions so the position in the practice set does not supply the decision.
For a learner ready for negative multipliers, compare −(x − 4) with −x − 4. The first equals −x + 4; the second is a different expression. This is an extension, not a compulsory next step for a child who is still learning positive multiplication over brackets.
Record what happened at the relevant step. If the expansion is now correct but arithmetic later fails, do not label the whole correction unsuccessful. The feedback may have addressed the expansion while another issue remains. Conversely, a correct final value with unexplained working does not establish why the learner succeeded.
Worked example: “Use the data” without inventing a conclusion
Imagine a fictional classroom exercise comparing two paper-helicopter designs. Three recorded fall times for design A are 1.2, 1.3 and 1.2 seconds; the times for design B are 1.5, 1.4 and 1.5 seconds. The exercise states that the drop height and release method were kept the same. These numbers are invented for this explanation, not results from an experiment we conducted.
A student writes, “B is better because its wings are longer.” A comment saying “Use the data” should lead the learner back to the actual question. If the question asks which tested design took longer to fall, “better” is too vague: better for what purpose? The wing difference alone does not report the observed outcome.
A more responsive answer is: “Design B took longer to fall in these trials. Its times were 1.4–1.5 seconds, compared with 1.2–1.3 seconds for A.” The answer identifies the measured difference and limits the claim to the trials shown.
A different question asking why a design falls differently would require an appropriate explanation, not just the same numerical comparison. The supplied times by themselves do not establish every part of that mechanism. Nor do they justify “longer wings always make every paper helicopter fall more slowly.”
The feedback lesson is to match evidence to the question’s demand. For the dedicated subject route, use How to Turn PSLE Science Feedback Into a Testable Repair. This general guide does not replace that subject-specific instruction.
Change the evidence, not just the numbers
A second fictional comparison might give A times of 1.2, 1.4 and 1.3 seconds, and B times of 1.3, 1.4 and 1.2 seconds. The two sets have the same average of 1.3 seconds. The learner should not carry the earlier conclusion that B is slower into a table that no longer supports it.
A useful response is, “These recorded results do not show a difference in average fall time between the two tested designs.” That does not prove that the designs can never differ. It states what this particular exercise supports.
This second task checks something the first correction did not: whether the student is reading the evidence afresh rather than copying the direction of the previous answer. Change can remove a previously justified conclusion as well as strengthen it.
When checking such work, keep the feedback target stable. The target is evidence-matched wording, not increasingly elaborate scientific language. A shorter, carefully limited answer can be a better use of feedback than a longer response with an unsupported claim.
Why explanations and fresh questions belong together
Butler, Godbole and Marsh compared correct-answer feedback with feedback that also explained the answer in two experiments using prose passages. On a final test two days later, the conditions performed similarly on repeated questions, while explanation feedback produced better performance on new inference questions. The finding concerns the tasks studied; it is not a guarantee that longer comments are always better. Research source: Butler, Godbole and Marsh, 2013.
For the worked examples in this guide, that distinction motivates a practical design choice: preserve a reason and reserve a new question. A corrected endpoint alone may be enough to repeat the old answer. A later task asks whether the learner can use what the explanation was intended to teach.
Do not confuse this with a requirement to explain every routine action at length. Select the explanation that addresses the actual error. The student who has misplaced one multiplication needs clarity about that relationship, not an unrelated lecture covering the entire algebra chapter.
Decide what to repair first when the page is crowded
A heavily marked assignment can contain several different jobs: a misunderstood question, missing evidence, weak structure, a spelling error and punctuation that interrupts meaning. Trying to act on every comment simultaneously may make it difficult to tell which change helped.
Start by identifying the change that is needed for the work to answer the question at all. If an essay argues about the wrong issue, improving adjectives is not the first priority. If a Mathematics model represents the wrong quantities, checking arithmetic alone cannot make it answer the intended question.
Then identify a manageable next target. The priority might be one paragraph’s evidence connection, one recurring expansion error or one qualification in a conclusion. This is not permission to ignore the remaining comments. Put them in a later queue and return once the first change can be demonstrated.
The number of active targets should depend on the learner and the task. “Exactly three targets” is no more universally correct than “correct everything immediately.” Choose a workload the learner can actually interpret and act on, then inspect the result.
Distinguish a missing explanation from missing knowledge
A student who cannot carry out a comment may not be refusing it. Consider “Explain why the denominator stays the same” after adding fractions with the same denominator. The learner may understand the requested action but lack the concept of a shared unit. More reminders to explain do not supply that knowledge.
Ask for a small demonstration. Can the learner represent the fractions, describe the unit and connect the representation to the operation? If not, instruction is needed before independent revision. Feedback points to the gap; teaching may be required to fill it.
The same distinction applies to writing. “Develop this argument” is difficult to carry out when the student does not yet know what counts as a reason, an example or an explanation. Offer a model, compare versions and then ask for an original attempt.
Keep a clear record of the support provided. Later, test what can be done with the intended aids available. Independence should never mean removing access arrangements or assistance that the learner legitimately needs to engage with the task.
What to do when two comments seem to conflict
“Explain more” and “Be concise” may appear contradictory. Sometimes they concern different weaknesses. The first may ask for the missing relationship between evidence and claim; the second may ask the student to remove repetition or irrelevant detail.
Return to the location of each comment. Suppose a paragraph says that reading is important, repeats that it is beneficial, then repeats that students should read. It may be long without explaining any particular benefit. A revision can remove the repetitions and add one clear mechanism. It becomes more explanatory and more concise at the same time.
Other disagreements are genuine. Two readers may prefer different stylistic choices, or the criteria may have changed between tasks. Do not pretend that every disagreement can be resolved by a universal checklist. Ask which criterion governs the present assignment and why a proposed revision serves it.
The student’s useful response is not automatic compliance with the latest voice. It is an evidence-based question: “Which part of this version answers the task less well, and what change would improve that?”
What to do when feedback appears incorrect
Feedback deserves attention, but it is not immune to error. A comment may refer to the wrong line, misunderstand a valid method or assume information not in the task. Treat a suspected error as something to investigate rather than as proof that all feedback is unreliable.
For a calculation, substitute the result, work through an alternative method or compare the expression with the original question. For a comprehension response, locate the textual evidence. For a factual explanation, consult an appropriate reliable source. Keep a distinction between “I disagree” and “I can show the specific problem.”
A respectful clarification might be: “I used this step because of this condition. When I substitute the answer, it satisfies the original equation. Could you show where the reasoning fails?” The request invites an explanation rather than a contest over authority.
If the feedback is corrected, preserve the change so the learner does not practise the mistaken version. If the teacher explains why the original comment was valid, return to the work and demonstrate the repaired understanding.
Use positive feedback to identify what to preserve
Learning from feedback is not only correcting wrong answers. A successful choice is worth understanding too. “This comparison is clear” can lead to a useful question: which feature made it clear? Perhaps the writer used the same criteria for both objects, named the difference precisely and avoided changing the basis halfway through.
Butler, Karpicke and Roediger’s 2008 experiments found that feedback improved later retention not only for initially wrong general-knowledge answers but also for correct answers given with low confidence. Their tasks involved multiple-choice responses followed by later cued recall, not whole school essays. The result is a reason not to assume that a correct response never needs informative confirmation. Research source: Butler, Karpicke and Roediger.
For practical revision, mark one successful decision to preserve alongside the weakness to repair. This protects a working part of the answer from unnecessary rewriting. The learner should be able to say both “I am changing this” and “I am keeping this because it serves the task.”
Make a revision explanation that shows the difference
After revising, ask for a brief explanation of what changed. It does not need to be an essay. “I added the two details that support nervousness and removed the invented background” is enough to make the intention visible in the English example.
For the algebra example, the explanation could be: “I originally multiplied only the first term. I have now multiplied both terms inside the bracket, then checked the solution in the original equation.” For the data example: “I replaced ‘better’ with the measured outcome and limited the claim to the trials shown.”
These explanations help distinguish purposeful revision from copying. They also give the teacher a quick way to notice a misunderstanding of the comment. If the learner says, “I made it longer because longer explanations get better marks,” the explanation itself reveals another issue to discuss.
Keep the explanation proportionate. The purpose is to show the decision, not to add an administrative task larger than the original correction. For younger children, pointing and explaining orally can be more useful than completing another form.
Retest the lesson after the model is no longer visible
An immediate revision answers, “Can the learner use this guidance now?” A later attempt answers a different question: “What remains available when the exact guidance and original solution are no longer beside the work?” Both observations are useful.
Roediger and Karpicke’s 2006 prose-learning experiments found an important timing contrast: restudying had an advantage on an immediate test, whereas retrieval practice produced better retention on delayed tests. Their experiments do not prescribe a universal school revision timetable. They demonstrate why immediate fluency and later availability should not be treated as identical outcomes. Research source: Roediger and Karpicke.
For this guide’s exercises, arrange a later fresh attempt within the learner’s actual timetable. That might be the next lesson or another suitable occasion. State whether a reminder was given, what aids were available and how the question differed. Do not invent a precise optimum interval for every age and subject.
If the lesson is forgotten, revisit the explanation and try again. The result tells you something about what needs support; it does not justify a judgement about the learner’s character.
A small feedback record that earns its place
A useful record can fit in a few lines: the task, the comment, the learner’s interpretation, the change made and the later check. Keep only information that helps someone make the next teaching or studying decision.
For the algebra case, an entry might read: “Bracket expansion; multiplied only the first term; now show the multiplier acting on both terms; revised original with help; fresh positive-bracket problem completed independently; negative multiplier not yet checked.” The last phrase matters. It prevents one narrow success from becoming a claim of complete mastery.
Record support as well as correctness. “Correct after a prompt” and “correct without a prompt” are different observations, though both can mark useful progress. A student should not be penalised for honestly recording the support that made the attempt possible.
Avoid creating a permanent catalogue of every small error. Review the record, move genuinely stable items into occasional maintenance and keep active attention on what remains useful. A feedback record is a working aid, not a character file or a public ranking system.
Build a short response session around actual work
A practical session can begin with one comment and one specific target. The learner identifies the relevant part of the original, paraphrases the feedback and makes a first revision. The teacher or parent checks whether the action matches the target before assigning more practice.
Next, ask for the revision explanation. If the learner can explain the change, use a fresh item or a new paragraph that requires the same decision. If not, clarify the comment or reteach the missing knowledge. The sequence should respond to the evidence rather than continue mechanically.
End by recording what was demonstrated and what remains to be checked later. The outcome might be a successful independent attempt, a clearer diagnosis or an agreed question for the teacher. Not every useful session must end with a perfect answer.
Allow the time needed for the task. A one-line expansion error and an essay’s argument structure do not require identical response periods. The essential feature is protected opportunity to act, not compliance with a universal twenty-minute routine.
Teachers: design the opportunity to use the comment
The Education Endowment Foundation’s 2021 feedback guidance highlights the need for strong initial teaching, feedback timed to move learning forward, and deliberate planning for how pupils will receive and use it. It emphasises principles rather than one mandatory written or verbal format. This is practice guidance, not evidence that the particular routine in this article guarantees a result. Source: EEF guidance announcement.
A practical implication is to prepare the response opportunity alongside the feedback. Before returning the work, decide where the learner will make a revision, what new task could check the same decision, and how you will notice whether clarification is needed.
A class does not need identical feedback targets simply because it completed the same assignment. One student may need to choose evidence; another may need to explain it; another may need to stop overclaiming. Use the work to group the response tasks where practical.
Do not solve every revision yourself to make the next version look successful. Provide enough instruction for the student to act, then inspect the student’s attempt. The purpose is better learning, not a folder of adult-polished pages.
Parents: make the conversation about the work
A useful opening is, “Show me one comment you understand and one you need help interpreting.” This avoids demanding that the child explain an entire paper while upset or tired. It also allows a parent to support the process without pretending to be an expert in every subject.
Ask the child to locate the passage or line being discussed. Then ask what would change in a new attempt. When the answer is unclear, help compose a short question for the teacher rather than filling the gap with a guess about the marking scheme.
Keep emotional support and subject correction distinct. A child can be disappointed about a result and still deserve a calm explanation of what the feedback means. There is no need to insist that criticism must feel pleasant before it can be discussed, nor to dismiss distress as an excuse.
Avoid turning a correction session into proof of obedience. The useful evidence is in what the learner understands and can do. A quiet nod is not a sufficient check, but neither is an articulate disagreement a failure to learn. Return to the task, the criterion and the demonstrated change.
Peer feedback without rewriting someone else’s work
Use a clearly stated purpose when students review one another’s work. In the invented English exercise, a partner can underline the claimed emotion, mark the textual evidence and ask where the connection is explained. That is a more focused task than “make your friend’s answer better.”
The author then decides what to revise and explains the decision. The partner should not quietly rewrite the whole answer. Keeping authorship with the learner preserves a chance to observe whether the feedback has been interpreted.
Peer comments may also need checking. A classmate can be confident and mistaken, or may offer a preference rather than a task-based reason. Bring unresolved disagreements to an example, the agreed criteria or the teacher.
For younger learners, begin with a narrow comparison using a teacher-prepared example. Asking children to judge everything about an unfamiliar genre may create confusion rather than useful feedback. The degree of responsibility should match what they have been taught to notice.
AI feedback: keep a record of what the learner actually changed
When using an AI tool, keep the learner’s original response and ask for comments before accepting a replacement. A bounded prompt might be: “Identify one place where my evidence does not support my claim. Explain the issue without rewriting the whole paragraph. Then ask me to revise it.” Follow the school’s rules on permitted assistance and avoid uploading identifiable student information unnecessarily.
After receiving the comment, verify it against the task and any relevant source. If the tool proposes a new factual claim or quotation, check it before including it. Do not treat confident wording as a substitute for a source or teacher judgement.
The learner should make the revision, explain it and attempt another task without copying the generated response. Record AI assistance honestly. A polished answer produced mainly by a tool cannot, on its own, show what the student can write independently.
The existing AI-Assisted Study guide provides the broader learning route. Here, the narrower question is whether the tool’s comment led to a learner-owned decision that can be demonstrated again.
When the learner keeps repeating the error
Check the sequence before increasing pressure. Was the feedback read accurately? Could the learner carry it out? Did the student revise independently or copy a model? Was the next problem similar enough to apply the lesson? Did the task contain additional demands that were not present before?
Use a small diagnostic item to separate possibilities. For the bracket error, ask for expansion without solving an equation. If expansion is secure alone but breaks inside a longer problem, the next support should address that larger context. Repeating the same explanation may not address the new difficulty.
If the original concept remains unclear, reteach it. If the learner cannot recognise when the rule applies, compare examples and non-examples. If the correction was only memorised, ask for the reason and a changed case. These are candidate responses to observed difficulties, not diagnoses that can be made from a score alone.
Also inspect the comment itself. It may be inaccurate, too broad or written in language the learner cannot yet use. Repeated failure is a reason to review the teaching exchange as well as the learner’s response.
Judge progress without inventing certainty
The strongest practical record combines several observations: the learner interprets the comment accurately, makes a relevant revision, explains the decision, and uses the lesson in a later fresh task with the intended level of support. These observations are more informative than counting corrections completed.
Comparisons still need care. A later task may be easier, more familiar or more heavily prompted. A changed score does not isolate the effect of feedback if extra teaching and practice also changed. Describe what improved without claiming more causal certainty than the situation allows.
For example: “The learner expanded all brackets correctly in this fresh four-item set without prompts” is a defensible observation when it actually happened. “This feedback method guarantees better Mathematics results” is not. In a real record, write the task, date and support conditions; the examples in this article are not measured outcomes.
A poor later attempt is also information. It may show a remaining weakness, a change in conditions or a need for another sample. Avoid replacing evidence with either automatic optimism or automatic pessimism.
A two-week trial, not a promise of transformation
A family or teacher can try the following arrangement with one manageable target. In the first week, retain the original work, clarify the feedback and complete a guided revision if needed. Include one fresh opportunity to use the same lesson and record what support was required.
In the second week, revisit the decision in another suitable task. Before showing the old comment, ask the learner to explain what they would check. Then inspect whether that check appears in the work. Finish by deciding whether to continue the target, change the support or move it into occasional review.
This is an illustrative timetable. It is not a prescribed dose, a guaranteed route to a particular grade or a substitute for subject teaching. If the first comment reveals a substantial knowledge gap, spend the time teaching it rather than forcing the plan to remain on schedule.
The useful outcome of the trial is a better-informed next decision. Keep a practice that the learner can use; adjust one that is not helping; retain uncertainty where the evidence is too thin to decide.
How this guide connects to the wider learning route
When the main problem is staying with difficulty before feedback arrives, return to How to Improve Productive Struggle. When the learner cannot formulate the question to ask a teacher, use How to Improve Help-Seeking.
When the difficulty is judging whether a strategy is working, use How to Improve Metacognition. When the task has ended and the learner needs to turn the experience into a next action, use How to Improve Reflection.
The How Learning Works mechanism map and Complete Learning Runtime Index provide broader navigation. Choose the route that matches the actual difficulty rather than trying to apply every study idea at once.
Research notes and limits
Carless and Boud, 2018: The development of student feedback literacy. This conceptual paper supplies the feedback-literacy framing. Applying that framing to the school examples here is an educational adaptation, not a claim that these examples were evaluated in that paper.
Butler, Karpicke and Roediger, 2008: Feedback and low-confidence correct responses. The experiments concern retention of general-knowledge responses. They support attention to uncertainty in correct responses, not a universal claim about every feedback format.
Butler, Godbole and Marsh, 2013: Explanation feedback and transfer of learning. The journal article appeared in 2013, following online-first publication in December 2012. Its comparison of repeated and new inference questions informs the distinction between fixing an answer and using an explanation elsewhere.
Roediger and Karpicke, 2006: Test-enhanced learning. This experimental work informs the distinction between immediate performance and later retention. It does not determine the ideal retest interval for an individual child.
Education Endowment Foundation, 2021: Teacher Feedback to Improve Pupil Learning: guidance announcement. This provides practice guidance on planning feedback and its use. No external source here validates an eduKate-branded guaranteed improvement formula, and no such formula is proposed.
The next answer is where the comment earns its value
Return to the paper that came home with “Explain your answer.” A useful response is not necessarily a longer correction. It may be one carefully selected detail, one explicit relationship, one repaired algebraic step or one qualification that stops a conclusion exceeding the evidence.
The learner should be able to show what changed and why. Then there should be another opportunity to use that decision without simply copying the original correction.
That is the practical aim: not more comments collected, more pages rewritten or more promises to be careful, but a learner who can recognise a relevant piece of advice and make better work because of it.
