The Tutor Handbook Vol No.0096 | The Correction Load Gate

The Tutor Handbook · Volume 0096 · Series ID THB-0096

The Tutor Handbook: Complete series index.

A marked page can become a strange kind of failure. The tutor has seen everything that is wrong, so the natural impulse is to say everything that is wrong. A calculation has a sign error, an inefficient method, missing units and a weak final statement. A comprehension answer uses broad evidence, over-explains one point and leaves another unsupported. An essay paragraph has a vague claim, an awkward sentence, repetitive vocabulary and a weak link between evidence and explanation. The tutor can identify all of it. That does not mean the learner can use all of it at once.

This is the problem of correction load. The educational job is not to maximise the number of defects mentioned. It is to decide which corrections deserve to become active work now, which should be noted but left dormant, which are consequences of an earlier error, and which can safely wait until the learner has enough control to act on them. A page covered in intelligent comments can still be instructionally poor if the learner cannot tell which one changes the next attempt.

The Correction Load Gate is the tutor’s decision about how much corrective information to activate from one performance, so feedback remains specific enough to improve the next attempt without becoming a second task that is larger than the learning problem itself.

This volume sits beside, but does not replace, How Feedback Works in Teaching and The Tutor Handbook Vol No.0072 | The Feedback Receipt. Those pages own the wider feedback mechanism and the question of whether feedback changed later action. This page owns a narrower tutor decision: when the performance contains many correctable features, how many should become live feedback now, and which ones?

Quick Answer

Effective correction is selective without being careless. The tutor first identifies the target of the task, separates root errors from downstream symptoms, protects correct thinking that should remain, and chooses the smallest set of corrections that can materially improve the next attempt. Other issues can be recorded for later sampling rather than activated simultaneously. The learner then uses the selected feedback on a fresh or revised attempt so the tutor can see whether the correction produced a better decision.

The gate matters because feedback has an opportunity cost. Every active correction uses learner attention, tutor explanation time, practice time and emotional bandwidth. The Education Endowment Foundation’s current feedback summary describes effective feedback as specific information that helps learners improve and stresses that feedback needs opportunities for action; it also warns that effects vary and can sometimes be negative. The Australian Education Research Organisation similarly frames progress monitoring as identifying gaps and responding with additional instruction, guidance or feedback where necessary. Neither source says “correct every visible error immediately.” The practical implication is to make correction answerable to the learning goal, the learner’s present state and the next opportunity to act.

1. What the Correction Load Gate owns

This volume owns prioritisation inside a correction event. Imagine a tutor reading one piece of work and noticing ten possible interventions. The question is not whether those observations are accurate. The question is which observations should become instructions, prompts, reteaching or practice right now.

The Correction Load Gate therefore begins after the tutor has enough evidence to see a problem and before the tutor releases a wall of comments. It can be used for spoken feedback, written marking, live correction, model-answer comparison, diagnostic review, homework return and post-paper analysis. It is especially useful when the work is rich enough to contain several interacting weaknesses.

It does not own the general science of feedback. It does not own the learner’s entire error history. It does not mean tutors should ignore serious misconceptions. It is not a fixed rule such as “give only two comments.” The correct load depends on the target, the learner’s knowledge, the cost of the errors, the dependencies among them and the amount of fresh practice available.

2. A learner does not receive comments; a learner receives a new task

A tutor may think the task is “read my feedback.” From the learner’s point of view, feedback creates a new performance demand. The learner must notice the comment, understand the language, locate the relevant part of the original attempt, reconstruct what went wrong, decide what to change, preserve what was already right and then execute the repair. Four comments can therefore represent four new learning tasks.

This is why comment count is a poor measure of feedback quality. Two well-chosen corrections can require substantial reasoning. Ten small proofreading marks may be cheap to process. One vague comment such as “develop more” can be harder to act on than three precise comments. Correction load is about the cognitive and decision burden created by feedback, not merely the number of annotations.

A useful tutor question is: What new decisions am I asking the learner to make because of this feedback? If the answer contains more decisions than the learner can realistically carry into the next attempt, the correction load is too high even if every comment is true.

3. Start with the task’s intended capability

Correction should be anchored to what the task was meant to reveal or practise. If the task was designed to test whether a learner can select a Mathematics method from an unlabelled problem, a beautifully written solution with one arithmetic slip may still provide strong evidence about method selection. If the tutor turns the entire review into arithmetic remediation, the correction event can drift away from the reason the task existed.

Likewise, if an English exercise is intended to test evidence selection, sentence-level polish matters but should not automatically displace the evidence problem. In Science, a learner may use imprecise language but show a correct causal chain; the tutor should distinguish the language repair from the mechanism being assessed. This does not mean form never matters. It means the correction order should respect the target before expanding to adjacent improvements.

The first gate question is therefore: What was this task for? If the tutor cannot answer, the learner will receive a mixture of comments without a hierarchy.

4. Separate root errors from downstream symptoms

Many pages look worse than they really are because one early error creates several later defects. A student misidentifies the percentage base, so the equation, calculation and final statement all become wrong. A reader chooses weak evidence, so the explanation becomes vague and the conclusion appears unsupported. A writer begins with a blurred claim, so later sentences repeat rather than develop.

Correcting every downstream symptom independently can waste learner attention. The higher-leverage move is often to repair the first decision that generated the rest, then ask the learner to reconstruct the later work. If the downstream errors disappear, they did not need separate interventions. If some remain, they become cleaner evidence for the next correction.

This is consistent with the broader Tutor Handbook principle of finding the first useful weak link, but the Correction Load Gate applies that logic inside one feedback episode. It asks which comment has the greatest chance of removing several other comments from the page.

5. Protect what is already right

Correction can accidentally erase correct reasoning. A tutor rewrites a paragraph so completely that the learner can no longer see which original decisions were good. A Mathematics solution is replaced by the tutor’s preferred elegant method even though the learner’s method was valid. A Science answer is converted into model language even though the learner’s causal structure was correct.

High-quality correction should identify success as well as error. The EEF feedback synthesis explicitly notes that feedback should not be reserved only for incorrect work. In practice, this means the learner needs to know what should remain stable while one part changes.

A useful pattern is: keep this, change this, then test it here. “Your evidence is relevant; keep it. The problem is that your explanation does not show how that evidence supports the inference. Add that link, then answer a fresh inference question.” The learner is not asked to rebuild everything. The correction isolates one live decision.

6. Distinguish must-fix, should-fix and can-wait

Not every visible problem has the same instructional priority. A tutor can sort corrections into three temporary classes.

Must-fix now: an error that blocks the target capability, creates a false concept, compromises safety or validity, or makes later practice misleading.
Should-fix soon: a real weakness that matters but does not need to compete with the current repair.
Can-wait: a lower-leverage issue, stylistic preference, rare edge case or surface flaw that does not currently justify learner attention.

These are not permanent labels. A “can-wait” vocabulary problem may later become the active bottleneck. A small notation issue may become must-fix when the learner moves into more advanced work. The purpose is to preserve sequence, not to declare some forms of learning unimportant.

For parents, this distinction can be reassuring. A tutor who does not correct every minor error is not necessarily overlooking it. The tutor may be protecting the learner’s ability to act on the correction that matters most.

7. Use error families instead of individual marks

A page can contain seven errors that belong to one family. An English learner repeatedly shifts tense in the same way. A Mathematics learner repeatedly drops a negative sign during substitution. A Science learner repeatedly states an observation where an explanation is required.

Marking every instance separately may create visual density without creating a better diagnosis. A stronger correction can name the family, show one or two representative instances, then ask the learner to find the remaining examples. That changes the learner’s role from receiving seven corrections to detecting one recurring pattern.

The tutor can say: “I have marked the first two places where the sign changes incorrectly. There are three more. Find them before I show you.” The learner now practises error detection as well as error repair. If the learner cannot find the remaining examples, that is useful evidence about the current support level.

8. Do not activate a correction that the learner cannot yet understand

Sometimes the tutor sees a sophisticated weakness before the learner has the conceptual language to work on it. A young writer’s paragraph has problems of rhetorical stance, but the learner is still struggling to make a clear claim and support it. A Mathematics learner’s solution could be more elegant, but the learner is still establishing a reliable valid method.

Advanced correction can be intellectually accurate and educationally premature. This is where tutor expertise can become over-teaching. The tutor notices more than the learner needs to carry today.

The gate therefore includes a readiness question: Does the learner have enough prior knowledge to act on this correction? If not, the tutor can record the issue, protect it from being forgotten, and wait until the prerequisite control exists.

9. Correcting the learner is different from editing the artefact

A polished artefact and an improved learner are not the same outcome. When the tutor edits a sentence into excellent English, the document improves. When the tutor replaces an inefficient Mathematics method with a cleaner one, the page improves. But the learner may have contributed very little to the transformation.

The Correction Load Gate should therefore favour feedback that leaves a meaningful operation for the learner. Instead of rewriting the whole paragraph, identify the sentence where the reasoning breaks and ask for a revision. Instead of completing all missing units, mark the first omission and ask the learner to audit the rest. Instead of correcting every scientific term, ask the learner to compare their explanation with a reference and identify where precision changes meaning.

The rule is not “never show the answer.” Sometimes direct explanation is necessary. The rule is that correction should be matched to the educational job. If the goal is learning, the learner eventually has to perform the repaired operation.

10. A composite Mathematics case: four errors, one live correction

The following case is fictional and constructed for teaching.

Alicia solves a rate problem. Her equation is correct, but she makes one arithmetic slip, omits units in two lines, uses an unusually long method and gives a final answer with the wrong unit because she copied it from the question. A tutor could produce four separate correction themes: arithmetic accuracy, units, efficiency and checking.

The target of the current lesson, however, is translating a word problem into a valid equation. Alicia did that correctly. The tutor first protects the success: “Your equation represents the relationship correctly.” Then the tutor asks Alicia to locate where the numerical answer first becomes inconsistent with her own equation. Alicia finds the arithmetic slip. The tutor does not yet insist on the elegant method because method efficiency is not the active target and her current route is valid.

The final-unit error is more important than the missing intermediate units because it changes the communicated answer. The tutor activates a second small correction: “Before you submit, read the question and state what the number represents.” Alicia repairs the final line. The missing intermediate units are recorded as a watch item for later work. One page contained four correctable issues; the active correction load was two.

11. A composite English case: do not correct the whole paragraph

Beatrice writes a comprehension explanation with a strong inference but weak evidence precision. The paragraph also contains repetitive vocabulary, one awkward sentence and an unnecessary introductory phrase. If the tutor marks all four features with equal visual weight, Beatrice may leave believing her “writing is bad.”

The active task is comprehension reasoning. The tutor underlines the selected evidence and asks, “Which words in this line do the real work for your inference?” Beatrice narrows the quotation and rewrites the explanation. The awkward sentence is corrected only if it prevents the reasoning from being understood. The repetitive word is left alone for now because lexical variety is not what the task is testing.

At the next composition lesson, vocabulary variety can become active. Correction is not omission. It is sequencing. Beatrice receives a clearer picture of what improved and why.

12. A composite Science case: one misconception outranks three wording issues

Ciara explains that condensation happens because “cold air enters the cup and turns into water.” Her response also uses an imprecise pronoun, misses a linking word and has a grammatical error. The conceptual model is wrong. That error deserves priority because leaving it untouched would allow later polished language to strengthen a false explanation.

The tutor does not spend the first minute correcting grammar. Instead, the tutor asks Ciara to identify where the water appears and where the water particles could have come from. A brief explanation or representation may be needed. After the concept is repaired, Ciara restates the explanation in her own words. Only then does the tutor address language features that prevent scientific precision.

This is a must-fix-now case. Correction load is selective, but selection does not mean avoiding difficult conceptual repair.

13. How to choose the first correction

When several issues compete, the tutor can use six questions.

Which error most directly affects the task’s target?
Which error sits furthest upstream in the reasoning chain?
Which error, if repaired, is likely to remove other errors?
Which error is likely to recur across future work?
Which correction can the learner act on now?
Which error would be costly or misleading to leave uncorrected?

The tutor does not need a numerical score. These are judgement prompts, not a validated algorithm. Their value is in forcing a comparison among possible corrections instead of treating every visible flaw as equally urgent.

If two errors remain equally important, choose the one with the clearer next attempt or split the work across two short cycles. The objective is readable improvement, not artificial simplicity.

14. How much feedback is too much?

There is no universal number. The same learner may handle different loads on different tasks. Three proofreading corrections can be lighter than one conceptual reconstruction. A senior learner may be able to work through several linked comments independently, while a younger learner may need one correction followed immediately by guided practice.

Instead of counting comments, inspect actionability. Ask whether the learner can state the first thing to change, why it matters, what remains unchanged and where to try it. If the learner cannot identify the next move because the comments compete, the load is too high.

Another signal is correction abandonment. The learner fixes the first two items and stops reading. Or the learner mechanically changes everything the tutor marked without being able to explain any of it. Those patterns suggest that feedback has become compliance work rather than learning.

15. Use a correction queue

Issues that matter but should not become active now can enter a small correction queue. The queue is not a permanent list of defects. It is a holding area for plausible next improvements.

Active now: evidence selection.
Next if active repair holds: explanation precision.
Maintenance: answer scope.
Watch only: vocabulary repetition.
Ignore unless recurrent: one punctuation slip.

At review, delete queue items that no longer appear. The learner should not carry an ever-growing record of past weaknesses. The queue exists to protect focus while preserving useful observation.

This connects to The Change Queue, but the scale is different. That volume orders learning-route changes. The Correction Load Gate orders feedback inside one piece of work or short performance cycle.

16. Feedback timing changes correction load

Immediate correction can reduce the learner’s need to hold an error state for long, but it can also interrupt productive thinking. Delayed feedback can preserve an independent attempt, but too much delay can make the original reasoning difficult to reconstruct. Existing eduKateSengkang work on feedback timing owns that broader mechanism.

For this volume, the practical point is that timing changes load. If five corrections are stored until the end of a ninety-minute tutorial, the learner may receive them as one dense package. If one conceptual correction is handled early and the learner gets a fresh attempt, later feedback can be based on the updated state.

A tutor can therefore reduce correction load not only by saying less, but by distributing correction across meaningful attempts.

17. Written marking should not become a private tutor monologue

Written comments are useful when the learner can interpret them later. They become weak when the page records the tutor’s intelligence more clearly than the learner’s next action.

Before writing a long comment, ask whether a shorter prompt would leave a better operation for the learner. “Why does this evidence prove the inference?” may be more useful than a paragraph explaining the entire answer. “Check the base” may be enough if the learner already understands percentage bases and simply selected the wrong one. If not, a terse comment can be cryptic and direct instruction is better.

The choice is not between detailed and brief feedback. It is between feedback that matches the learner’s current knowledge and feedback that does not.

18. Spoken correction should preserve thinking time

Live tutoring makes it easy to correct too quickly. The tutor sees the mistake, knows the fix and fills the silence. The learner receives a smooth explanation but loses the chance to locate the break.

Stanford’s National Student Support Accelerator emphasises open-ended questions, student voice and productive struggle in tutoring. Its small-group guidance also encourages tutors to become comfortable with silence rather than immediately jumping in. The point is not that silence is always beneficial. The point is that a correction can begin with a question that reveals whether the learner can self-locate the error before the tutor supplies the fix.

A simple sequence is: “Show me where you became least certain.” Then listen. If the learner identifies the correct region, the tutor can narrow support. If the learner is lost earlier, the correction moves upstream.

19. Three-student tutorials: correction load is individual even when feedback is shared

In a three-student tutorial, one learner’s error can become a useful shared case. But group correction creates a special load problem: two students may receive corrections they do not need, while the student who made the error may feel exposed.

The tutor should share a correction only when the underlying learning job is genuinely useful to the group. A method-selection misconception that all three are likely to encounter may justify a short shared discussion. A personal pattern of punctuation slips usually does not.

When sharing, protect provenance. “Here are two possible solutions; which one preserves the condition?” is often better than “Alicia made this mistake.” The learner’s work is not public property merely because the group is small.

20. Parent communication: fewer comments can mean better teaching

Parents often use visible marking as a proxy for tutor effort. A heavily annotated page looks thorough. A lightly marked page can look careless. Tutors should explain the difference between documenting every issue and activating the right correction.

A useful parent update is: “I saw four issues in this paper. We are actively repairing evidence selection because it is the earliest recurring weakness. I have recorded the vocabulary and sentence issues, but I am not asking her to carry all of them at once. Once the evidence problem holds on fresh passages, we will see what remains.”

This communicates professional selectivity. It also helps parents avoid adding a second correction layer at home. If the tutor is focusing on one repair while the parent adds five more, the learner experiences the combined load, not the adults’ separate intentions.

21. Learner communication: make the correction hierarchy visible

The learner should know what matters now. “There are several things we can improve, but today we are fixing one: how you connect the evidence to the inference. Your evidence choice is mostly strong, so keep that. We will revisit sentence polish later.”

This language protects agency and avoids global self-judgement. The learner is not “bad at English.” One decision is under repair. Other aspects are stable or waiting.

As learners mature, invite them into the prioritisation. Ask: “If you could fix one thing on this page that would improve the whole answer, what would you choose?” Their choice may be wrong, but the discussion exposes their model of improvement. Over time, correction prioritisation can become part of self-regulation.

22. When the learner wants every error corrected

Some learners feel safer when every visible error is marked. This preference should be taken seriously but not obeyed mechanically. Ask why. The learner may fear missing something before an examination, may be accustomed to exhaustive marking, or may not trust their own error detection.

One compromise is to separate audit from active correction. The tutor can indicate that additional issues exist—perhaps with a light symbol or end note—while highlighting the one or two active repairs. Later, the learner can return to the audit list if the main correction is stable.

The goal is not to hide information. It is to preserve a hierarchy between what has been noticed and what the learner is expected to change now.

23. When exhaustive correction is appropriate

Selective correction is not a universal virtue. Some tasks are explicitly editing tasks. A final submission may require a comprehensive accuracy sweep. Safety-critical or rule-critical work may demand immediate correction of every consequential error. A learner preparing a final examination response may need to know all remaining mark-losing features even if not all can be repaired deeply in one session.

The gate therefore asks about purpose and timing. Early learning often benefits from narrow, high-leverage repair. Final quality assurance may need broader coverage. The tutor should tell the learner which mode they are in.

“Today we are learning from this draft” is different from “today we are auditing this final draft.” Confusing those modes produces either under-correction or overload.

24. Research foundation and limits

The research base supports several principles that inform this volume, but it does not validate a universal Tutor Handbook “correction load” protocol. The EEF’s feedback synthesis reports a strong overall evidence base while also noting wide variation in effects, the importance of task- and process-focused feedback, opportunities to act on feedback and workload/opportunity-cost considerations. Its summary was last formally reviewed in 2021 even though the Toolkit remains live, so readers should not treat the page as a 2026 systematic review of every new study.

AERO’s Monitor Progress guidance, last updated 14 May 2026, recommends frequent checks for understanding and responsive additional instruction, guidance or feedback. Its examples show teachers responding to misconceptions and monitoring errors, not merely accumulating marks. OECD’s 2025 Unlocking High-Quality Teaching describes formative assessment as an ongoing process of setting goals, diagnosing learning, giving feedback and adapting to student thinking, while acknowledging that timing and individual needs are complex.

These sources support targeted, actionable, responsive feedback. The specific prioritisation rules in this article are an educational design proposal for tutoring, not a validated clinical or psychometric instrument. Tutors should adapt them to learner age, subject, task, access needs and real school requirements.

25. The Correction Load Gate card

What capability was this task meant to reveal or practise?
What did the learner already do correctly that should remain stable?
Which error is earliest in the causal chain?
Which correction would remove several downstream symptoms?
Which errors are recurring rather than one-off?
Which corrections can the learner actually understand and act on now?
Which issue would be costly to leave uncorrected?
What can wait without creating harm or false learning?
What fresh or revised attempt will let the learner use the active feedback?
How will I know the learner changed, rather than only the page?

If the tutor cannot identify a clear next attempt, the feedback package probably contains more commentary than instruction.

26. The ethical standard

Correction is an exercise of power. The tutor controls what becomes visible, what is named as wrong, what is preserved and what the learner must revisit. Exhaustive correction can make the learner feel that every performance is a field of defects. Minimal correction can hide important gaps. The ethical position is neither harshness nor softness. It is proportion.

A professional tutor should tell the truth about the work without making the learner carry every truth simultaneously. The tutor should not conceal consequential misconceptions. The tutor should not inflate a page with corrections that mainly display expertise. The tutor should preserve legitimate access supports, distinguish stylistic preference from genuine error, and give the learner a real opportunity to perform the repaired operation.

The best correction is not the page with the most tutor ink. It is the smallest truthful set of feedback that helps the learner make a better next decision, then proves that the learner can carry that decision forward.

Evidence and connected reading

Final compression

A tutor can see ten things wrong and still choose two corrections. That is not lower standards. It is sequence.

Anchor correction to the task. Protect what is right. Find the upstream error. Distinguish must-fix from should-fix and can-wait. Group repeated mistakes into error families. Do not activate a correction the learner cannot yet understand. Leave a meaningful operation for the learner. Use a correction queue for what matters later. Match feedback timing to the attempt. Separate learning mode from final-audit mode. Then ask for a fresh receipt.

The objective is not a cleaner page. It is a learner who can make a cleaner decision without the tutor standing beside them.

That is the Correction Load Gate.

That is Tutor Handbook Volume 0096.

Small Group Tutorials

The Tutor Handbook Vol No.0096 | The Correction Load Gate — How a Tutor Chooses Which Errors to Correct Now Without Turning One Piece of Work Into an Unusable Wall of Feedback