PSLE English Comprehension Annotation | Mark Less, Remember More and Re-enter the Passage Faster

Annotation as an information interface

A passage is already an information interface. Paragraphs, punctuation, sentence order and typography organise meaning before the student touches the page. Annotation adds a second interface layer. That second layer should improve navigation and memory without competing with the first.

This perspective changes the question from “What should I underline?” to “What information will I need to retrieve later, and what is the cheapest visible cue that will help me retrieve it?”

The best annotation therefore behaves like a good interface: sparse, predictable, legible and task-sensitive.

Information scent

A useful mark gives the student a strong clue about what lies nearby. A margin label “result” has good information scent because it tells the reader what kind of meaning is in that paragraph. A star with no defined meaning has weak information scent.

When returning after several questions, the student should be able to glance at a mark and know why it deserves attention. If the mark requires reconstructing the original thought, it has poor retrieval value.

Retrieval cue quality

High-quality retrieval cues are:

• specific enough to distinguish one passage region from another;
• short enough to recognise instantly;
• stable in meaning across passages;
• attached to information likely to be revisited;
• unlikely to be mistaken for certainty when it merely indicates relevance.

“P4 result” is usually a stronger cue than “important”. “?” is a stronger cue for unresolved meaning than a star if the student has defined it consistently.

Cue specificity

A cue can be too broad. “Problem” beside three paragraphs does not help if the passage contains several problems. A cue can also be too narrow. Writing the full answer proposition in the margin duplicates work and may lock the interpretation too early.

Good specificity identifies the retrieval category without replacing the later thinking.

Cue stability

Students benefit when the same mark means the same thing over time. If “?” means unresolved, keep it that way. If “T” means turning point, do not use T for topic on another passage. Stable cues reduce decoding cost.

Cue competition

Marks compete visually. Ten equally dark underlines create ten equally strong signals. A useful annotation system uses contrast sparingly so the eye can find priority quickly.

This does not require different colours. It can be achieved simply by marking fewer things.

The salience budget

Every page has a limited salience budget: only a small number of items can stand out strongly at once. Annotation spends that budget.

If the student spends salience on obvious character names, descriptive adjectives and every connector, there is less visual distinction left for the real turning point or qualifier.

Salience should follow expected future value

The most visually noticeable marks should correspond to information most likely to matter later: passage update, high-value qualifier, central contrast, difficult return point or evidence corridor.

Annotation lifecycle

A mark has a lifecycle. It is created for a reason, used, possibly updated, and eventually becomes irrelevant or internalised.

• Create: a future retrieval or boundary risk is identified.
• Use: the mark routes reading, evidence search or verification.
• Update: later evidence changes what the mark should mean.
• Clear mentally: the mark’s job is complete for the current question.
• Retire from training: the student no longer needs the visible scaffold.

Many annotation systems teach creation but never teach retirement. The page gradually accumulates old signals without a rule for when they stop mattering.

Created mark, unused mark

If a student repeatedly creates a category of annotation and almost never uses it later, the category deserves review. A mark that looks intellectually respectable but has no downstream function is operationally weak.

Created mark, harmful mark

A mark can be worse than useless if it biases interpretation. Boxing “jealous” before evidence is secure gives the word more status than it deserves. A question mark would have been safer.

Used mark, high return

A tiny “O” beside the outcome paragraph is used three times for Questions 6, 8 and 10. It cost one second to create and saved several searches. This is high-return annotation.

Used mark, low return

A full-sentence margin summary is used once and saves only a brief reread. The information was helpful, but the representation may have been too expensive.

Annotation as a compression layer

Good annotation compresses meaning without replacing it. “P4 = correction” can stand in for a paragraph describing how a failed first attempt was modified. The student can later re-enter P4 and recover details from the original text.

The mark should therefore preserve where and what kind of meaning exists, while the passage retains the full content.

Compression failure: mark too vague

“Important” compresses too aggressively because it does not say what kind of importance. Re-entry still requires scanning the region from scratch.

Compression failure: mark too complete

Writing “The outdated notice caused residents to miss the new hours, which explains low attendance” duplicates the paragraph’s causal proposition. It may be useful in teaching, but it is too expensive as a routine exam mark.

The minimum useful cue

Ask what is the shortest mark that will reliably reconstruct the needed relationship later.

“old notice → unaware”

That may be enough to recover the larger causal paragraph when needed.

Dual indexing: question side and passage side

One of the strongest annotation systems uses two light indexes.

• Question side: preserve the semantic contract—quantity, actor, time, relation, source.
• Passage side: preserve the evidence location or structure needed to fulfil that contract.

The two indexes reduce different risks. The question-side mark protects relevance. The passage-side mark protects retrieval.

Dual-index example: two reasons

Question-side mark: circle two. Passage-side marks: one small tick at each confirmed reason. The student can see both the obligation and the evidence coverage without writing a full plan.

Dual-index example: before and after

Question-side mark: bracket how did her behaviour change. Passage-side marks: one light underline on initial avoidance, one on later volunteering.

Dual-index example: source condition

Question-side mark: box according to the survey. Passage-side mark: “S” beside the survey finding. The student is less likely to answer from later organiser commentary.

Dual-index failure: correct passage, wrong question contract

The student finds the turning point perfectly but forgets that the question asks for two effects. Passage annotation works; question annotation fails.

Dual-index failure: correct question contract, weak passage route

The student knows two reasons are needed but repeatedly scans the whole passage. Question annotation works; passage indexing remains weak.

Question-number indexing as a temporary bridge

Writing a tiny Q-number beside evidence can be helpful for students whose search cost remains high. It creates a direct route from answer space back to text.

But Q-number indexing should usually happen after evidence ownership is known. Premature numbering is dangerous because the mark can harden an incorrect route.

Question-number indexing and shared evidence

One passage zone may support Q3 and Q7. Two small numbers can coexist beside the same region if the evidence genuinely serves both. Shared evidence does not mean the answers should be identical.

Question-number indexing and distributed evidence

A question may need evidence from P1 and P6. The same Q-number can appear in two regions, reminding the student that the answer is distributed rather than local.

Question-number indexing should fade

Once the student can retain evidence ownership without repeated search, visible question numbers may no longer justify their cost. They are scaffolds, not identity.

Annotation and evidence corridors

An evidence corridor is the smallest passage span the student needs to revisit for one question. It can be a clause, sentence, sentence pair or two separated anchors.

Annotation can make the corridor visible without pre-writing the answer.

Corridor precision

Too narrow: one clue is marked when a second clue is necessary. Too wide: the whole paragraph is bracketed when only two clauses matter. The corridor should match the proof topology from PCEP-010.

Corridor marks and answer construction

Once the corridor is visible, the student extracts the required proposition. The bracket is not permission to copy everything inside it.

Annotation and evidence status

A useful advanced distinction is between three visible statuses:

• Anchor: useful place to return.
• Candidate clue: relevant but not yet sufficient.
• Confirmed evidence: evidence ownership is established for the current question.

The student does not need three different symbols. Even one consistent convention—such as adding the Q-number only after confirmation—can separate candidate from confirmed status.

Annotation should not simulate certainty

Boxes, dark circles and heavy underlining can psychologically convert a candidate clue into “the answer”. Use light provisional marks for uncertain material. Visual force should roughly match evidential confidence.

Mark intensity

Some students naturally vary mark intensity: a light underline for possible relevance, a tiny question mark for uncertainty, a brief margin label for structure. This can work if the distinctions are automatic. It should not become another complex code to learn.

Stale marks and visual version control

When the passage model changes, annotations should reflect the latest version. An obsolete arrow or boxed motive can remain visually powerful even after the student has mentally changed interpretation.

Cross out, amend or supersede stale marks clearly enough that the latest model is visually dominant.

Version-control example

low interest ? → low interest awareness problem

The old hypothesis remains historically visible but no longer controls later reasoning.

Do not create an archaeology problem

If several layers of corrections, arrows and cross-outs accumulate, the page can become difficult to interpret. A late model update may be easier to represent with one fresh margin note than by repeatedly editing every earlier mark.

Annotation after a difficult question

When one question creates uncertainty, students sometimes add more annotations across the passage to regain control. This can spread the disturbance.

A better recovery is local: mark the unresolved evidence zone or question, then reset to the next item using the existing passage anchors.

Annotation and cognitive closure

Question-number anchors can sometimes help closure. Once the answer is complete and verified, the student mentally treats the Q-number link as done. It does not need to be reread unless the answer reopens.

Annotation should not keep every question open

A page covered with Q-numbers can tempt the student to revisit everything at the end. Final review should still follow PCEP-011’s risk hierarchy, not annotation density.

Question marks on the question itself

Students often think annotation refers only to the passage. In many cases the highest-value mark belongs on the question because it preserves the contract while the student moves attention away from it.

A single circle around another, only after, according to the survey or the named actor may prevent an entire wrong answer.

The contract-return habit

After collecting evidence, glance once at the marked contract before writing. This short return aligns proof with obligation and reduces broad-topic answers.

Contract annotation can be personal

Maren may circle quantity. Iona may underline certainty language such as suggests. Leonie may mark time boundaries late in the paper. The visible cue should target the student’s actual risk.

Annotation and handwriting economy

Margin notes should be short enough not to compete with answer writing. Abbreviations can be useful when instantly understood. Long notes create handwriting cost that may have little return.

Abbreviation safety

A private abbreviation used only as a retrieval cue can be extremely short. It does not need formal grammar. But if the abbreviation is ambiguous to the student later, the saved writing time is lost during decoding.

Annotation density by passage region

Not every paragraph deserves equal marking. Turning points, qualifications and high-value evidence zones may be denser. Descriptive setup can remain nearly unmarked. Uneven annotation is often a sign of good prioritisation.

Uniform density can signal ritual

If every paragraph contains exactly two underlines and one margin note regardless of structure, the student may be following a procedure rather than reading meaning.

Annotation density by question type

A direct detail question may need no new passage annotation. A distributed change question may justify two anchors. An evaluation may need criterion and outcome marks. Annotation volume should respond to proof structure.

Annotation density by confidence

Low confidence should not automatically increase marking. First identify what is uncertain. If the uncertainty is local and targetable, one mark may be enough. General low confidence combined with broad annotation can become a reassurance loop.

Annotation and re-reading behaviour

One of the best measures of annotation quality is how it changes rereading. Strong annotation should shift rereading from global to targeted.

• Before: reread whole paragraph to remember its role.
• After: margin label restores role; reread only exact evidence sentence.
• Before: reread from passage beginning after disruption.
• After: T and O anchors restore the model immediately.

Annotation and route accuracy

If marks make the student return quickly to the wrong region, the system is efficient but incorrect. Measure both speed and route accuracy. False anchors must be repaired before they become automatic.

Annotation and score stability

Useful annotation should reduce variance across unfamiliar passages by protecting structure and scope. If a student’s system works only on familiar narrative passages but becomes cluttered on informational texts, transfer remains incomplete.

Annotation under unfamiliar topic load

When vocabulary is unfamiliar, students may mark more simply because the page feels harder. This is a poor trigger. Marks should respond to structural risk, not emotional difficulty.

If an unfamiliar term controls the cause or comparison, mark it. If it is decorative and the sentence’s role remains clear, leave it alone.

Annotation under fatigue

Late in the paper, the student may have less patience for annotation. The system should degrade gracefully: preserve question contract and the most valuable passage anchor, abandon low-value labels.

Annotation under haste

Haste can create sloppy false anchors. Under severe time pressure, it may be safer to mark the question qualifier and route mentally than to create hurried passage notes whose meaning is uncertain.

Annotation under overconfidence

Fast confident readers may skip all marks even when a known risk trigger appears. Selective annotation is not only for weak readers; one boundary mark can protect a strong reader from premature closure.

Original annotation case 61: strong information scent

A margin note “result” immediately tells Leonie why P6 matters when she returns later. She re-enters the passage at the correct place within seconds.

Original annotation case 62: weak information scent

A star beside P6 tells her only that she once thought something there was important. She must reread the paragraph to rediscover why.

Original annotation case 63: cue too broad

Three paragraphs are labelled “problem”. The student cannot distinguish original problem, implementation problem and later side effect.

Original annotation case 64: cue fitted

The labels become “access problem”, “notice problem” and “cost side effect”. Retrieval categories are short but discriminating.

Original annotation case 65: cue too complete

The student writes an eighteen-word causal explanation in the margin. The note is accurate but takes as long to reread as the original two sentences.

Original annotation case 66: stable symbol

“?” has meant unresolved for months. Iona sees it and immediately knows not to treat the adjacent clue as proven.

Original annotation case 67: unstable symbol

“?” sometimes means difficult, sometimes unknown vocabulary, sometimes unresolved inference. The mark cannot guide a specific future action.

Original annotation case 68: salience spent well

Only one phrase in the passage is boxed: “only after”. It genuinely controls a later cause question.

Original annotation case 69: salience spent badly

Five descriptive phrases are heavily underlined because they sounded interesting. The decisive qualifier receives no special attention.

Original annotation case 70: lifecycle completed

Q4 is linked to one evidence corridor, the answer is constructed and verified, and the number is mentally closed. The mark remains on the page but no longer occupies attention.

Original annotation case 71: lifecycle never closes

Every Q-number remains visually active in Leonie’s mind. Final review becomes a tour of all marks rather than a risk-based process.

Original annotation case 72: unused category

The student circles every connector but almost never uses those circles to answer or navigate. The category should be retired unless a specific connector error emerges.

Original annotation case 73: high-return category

A small paragraph-role label repeatedly prevents global rereading. The category remains valuable.

Original annotation case 74: minimum useful cue

“old notice → unaware” is enough to restore the causal mechanism. The student does not need a full-sentence note.

Original annotation case 75: compression too far

The student writes only “notice”. Later she cannot remember whether the notice was cause, solution or outcome. The cue lost functional information.

Original annotation case 76: dual index works

Question has “two reasons” circled. Two passage reasons receive tiny ticks. Maren writes both without rereading the question twice.

Original annotation case 77: question-side only

“Two” is circled, but evidence locations are not retained. Maren knows the quantity but rereads broadly to recover the reasons.

Original annotation case 78: passage-side only

Two evidence lines are underlined, but the quantity word in the question is forgotten. Only one line enters the final answer.

Original annotation case 79: shared evidence indexed safely

P4 supports Q3’s cause and Q7’s evidence-of-change question. Both numbers sit beside the paragraph, but the student constructs different answer relations.

Original annotation case 80: shared evidence creates copied answer

The same P4 mark leads the student to reuse Q3’s answer wording in Q7. Evidence was reusable; the answer proposition was not.

Original annotation case 81: distributed evidence indexed clearly

Q8 appears beside P1 goal and P6 outcome. The duplicated number reminds the student that evaluation requires both regions.

Original annotation case 82: distributed indexing overused

Q8 is written beside five paragraphs “just in case”. The index no longer narrows the search.

Original annotation case 83: corridor fitted

A bracket covers two sentences that jointly establish reluctance. The student returns to exactly that span for an inference.

Original annotation case 84: corridor undercovers

Only the final behaviour is bracketed for a change question. The initial state remains unindexed and is omitted.

Original annotation case 85: corridor overcovers

The entire six-line paragraph is boxed. Re-entry is faster than a whole-passage reread but still slower than a fitted two-sentence corridor.

Original annotation case 86: candidate clue stays provisional

A light underline marks one possible motive clue. Only after a second discriminating clue appears does the student add the question number.

Original annotation case 87: candidate becomes visually certain too early

The first clue is boxed and labelled Q5 immediately. Later counterevidence struggles to dislodge the anchor.

Original annotation case 88: update mark clear

The student crosses one outdated arrow and writes a new margin label. Later questions route through the revised model.

Original annotation case 89: update creates archaeology

Three layers of arrows and notes overlap. Leonie can no longer tell which is current. One clean superseding note would have been better.

Original annotation case 90: disruption handled locally

Question 5 remains uncertain. The student marks only that question and evidence zone, then uses existing T/O anchors to resume Q6.

Original annotation case 91: disruption spreads globally

The difficult Q5 causes the student to add new marks throughout the passage. The visual model becomes denser exactly when attention is already strained.

Original annotation case 92: contract-return habit works

After evidence collection, Maren glances back at circled “another reason” and realises her second clue repeats the first. She searches for a distinct reason.

Original annotation case 93: contract mark ignored

“According to the survey” is boxed, but the student never returns to the question after locating evidence. The mark exists without use.

Original annotation case 94: abbreviation automatic

“qual.” instantly means qualification to Iona. She can re-enter the reasoning without decoding delay.

Original annotation case 95: abbreviation ambiguous

“R” could mean reason, result or reader. The saved writing time is lost when the student hesitates later.

Original annotation case 96: uneven density is intelligent

P1 setup is unmarked. P4 turning point has two light marks. P6 qualification has one. The pattern reflects structural value, not neglect.

Original annotation case 97: uniform density signals ritual

Every paragraph gets exactly one underline and one margin word. The method ignores meaning differences.

Original annotation case 98: low-confidence target identified

Iona is unsure whether “they” refers to residents or organisers. She marks the pronoun and antecedent zone only. Uncertainty becomes local.

Original annotation case 99: low confidence creates broad marking

The entire paragraph is underlined because the student “is not sure”. No information target is established.

Original annotation case 100: rereading becomes targeted

Question asks why the second attempt worked. The student uses “correction” and “result” labels to return directly to P4–P5 instead of rereading P1–P5.

Original annotation case 101: false anchor speeds wrong route

A star beside P2 makes the student return there quickly for every cause question, even though Q7’s cause is in P4. Fast wrong routing is not a success.

Original annotation case 102: unfamiliar topic does not inflate marks

A science passage contains several unfamiliar nouns. The student marks only the one term that controls the comparison. Passage structure remains visually clean.

Original annotation case 103: unfamiliar topic triggers saturation

Every unfamiliar noun is circled. The page feels even more difficult because the marks amplify uncertainty.

Original annotation case 104: late-paper compression works

With less time, Leonie marks only the question qualifier and one evidence corridor. Lower-value paragraph labels disappear without accuracy loss.

Original annotation case 105: late-paper annotation collapses completely

Under time pressure, all marks stop. A familiar two-part risk returns and one slot is omitted. The system did not degrade gracefully.

Original annotation case 106: haste avoids false marks

Rather than drawing a hurried cause arrow, the student marks only “why” in the question and reads the evidence relation mentally. Minimal marking is safer under the state.

Original annotation case 107: haste creates false relation

A rushed arrow between adjacent events later causes a sequence-as-cause error. Annotation speed without meaning is dangerous.

Original annotation case 108: strong reader uses one boundary mark

Iona understands a paragraph instantly but underlines “may” because the evaluation question is likely to tempt overclaim. One mark protects an exceptional risk.

Original annotation case 109: strong reader refuses all annotation

A long multi-character passage later produces actor drift. Zero annotation became dogma rather than efficiency.

Original annotation case 110: developing reader uses explicit roles

P1 setup, P2 problem, P3 attempt and P4 outcome are labelled. The visible structure allows successful routing while the internal model is still developing.

Original annotation case 111: developing reader never fades scaffolds

Months later, the student still writes full role labels on every paragraph and cannot work efficiently without them. Training did not include compression.

Original annotation case 112: Q-number helps final review

A changed Q8 answer is easy to verify because the evidence region is already indexed. One targeted check resolves the risk.

Original annotation case 113: Q-number causes unnecessary final review

The student treats every numbered passage region as something that must be revisited at the end. Indexing becomes debt.

Original annotation case 114: criterion mark used correctly

“goal” in P1 is paired with “result” in P6. Evaluation asks whether the programme worked. The two anchors recreate the criterion–outcome structure immediately.

Original annotation case 115: criterion mark overgeneralised

A “goal” mark makes the student assume every later positive outcome proves success, even when the outcome is on a different dimension.

Original annotation case 116: counterevidence given equal status

A small “but” bracket beside the limiting sentence prevents three earlier supporting clues from dominating the final inference.

Original annotation case 117: counterevidence visually buried

The contradiction is lightly marked beneath a dense set of boxes. The visual hierarchy still favours the old interpretation.

Original annotation case 118: annotation category retired appropriately

Pronoun-reference marks have not changed performance for six weeks because actor tracking is automatic. The visible category is removed.

Original annotation case 119: retired category reactivated by new risk

A new passage with four same-gender characters causes confusion. Temporary entity anchoring returns because the structural risk has changed.

Original annotation case 120: interface succeeds under pressure

The student’s page contains a handful of meaningful marks: one question qualifier, two paragraph-role anchors, one uncertainty sign and one question-number evidence corridor. Each mark has a future job. None requires translation. After a difficult item, the student re-enters the model immediately. Annotation is doing exactly enough.

Annotation drill 21: information-scent audit

Cover the passage text but leave margin marks visible. Ask what each mark tells the student about the hidden region. Weak cues are revised.

Drill 22: salience budget

Allow only two high-salience marks on one passage. Students must decide what deserves strongest visual priority.

Drill 23: annotation lifecycle audit

For every mark in one passage, record create, use and final status. Identify unused, stale and retired marks.

Drill 24: minimum-cue compression

Shorten full-sentence notes until further compression would make later retrieval unreliable.

Drill 25: dual-index training

Use one semantic mark on the question and one or two passage anchors. Compare completeness with passage-only annotation.

Drill 26: shared-evidence indexing

Place two question numbers beside one evidence zone, then require different answer propositions. This separates evidence reuse from answer reuse.

Drill 27: distributed-evidence indexing

Use the same Q-number in two separated passage zones. Students build an answer that integrates both without rereading the whole text.

Drill 28: candidate-to-confirmed transition

Use a light provisional mark first. Add a Q-number only after evidence ownership is confirmed. Compare with immediate boxing.

Drill 29: version control

Give a passage where later evidence overturns the first interpretation. Students must revise old marks so the final model is visually unambiguous.

Drill 30: archaeology cleanup

Present an over-edited page. Students create one clean superseding annotation rather than adding another layer of arrows.

Drill 31: contract-return habit

After collecting evidence, students must glance at the marked question boundary once before writing. Measure scope errors.

Drill 32: density-by-region audit

Compare annotation density with paragraph function. Ask whether dense marking corresponds to high structural value or merely difficult wording.

Drill 33: reread transformation

Count global and targeted rereads before and after a new annotation system. The desired change is fewer global rereads, not necessarily fewer rereads overall.

Drill 34: route-accuracy check

Record whether each mark routes to the correct evidence zone. Fast false anchors are identified and corrected.

Drill 35: unfamiliar-topic restraint

Use a passage with difficult vocabulary and impose a strict annotation budget. Marks must target structure rather than emotional unfamiliarity.

Drill 36: late-paper compression

Repeat the same passage strategy with less time. Identify which marks disappear first and whether high-value boundaries remain protected.

Drill 37: overconfidence trigger

Use a passage that looks easy but contains one subtle exception or chronology shift. Strong readers must decide whether one mark would prevent premature closure.

Drill 38: delayed re-entry test

Return to the annotated passage after a longer unrelated task. Measure how quickly the marks reconstruct structure.

Drill 39: category retirement

Remove one established annotation category for several passages. Restore it only if a meaningful error returns.

Drill 40: interface stress test

Use an unseen passage under realistic timing with one interruption. Evaluate not how much was marked, but whether every mark had information scent, survived update, supported re-entry and stayed affordable.

The second-layer annotation standard

At this level, annotation is no longer a collection of study tips. It is a designed interface between the student and a passage that must be revisited repeatedly under time.

Maren uses the question side to protect scope and the passage side to protect retrieval. Iona uses provisional marks and visible version control to keep interpretations flexible. Leonie uses sparse high-information cues so re-entry remains cheaper than rereading.

The mark is successful when it disappears into the performance system: seen instantly, used once or twice, then mentally cleared.

The mark-or-do-not-mark decision

The final annotation skill is not knowing many ways to mark a page. It is deciding whether a mark is worth creating at all.

Every proposed annotation can pass through a compact threshold:

• Will I probably need this information again?
• Would I otherwise lose, confuse or spend time relocating it?
• Can one small mark protect it?
• Will the mark remain instantly interpretable later?
• Is the mark less costly than the rereading or error it is likely to prevent?

If several answers are no, the page may be better left untouched.

The annotation threshold should rise as reading becomes stronger

Early in training, students may mark many relationships because those relationships are not yet automatically represented. Later, the same student can hold paragraph roles, actor ownership and basic chronology internally. The threshold for external marking should rise.

This is a sign of learning. A more capable reader should generally need fewer visible supports for ordinary structures and reserve annotation for exceptional risk.

The annotation threshold can also fall temporarily

A difficult passage, unusual chronology, dense set of similar actors or unfamiliar informational structure may justify more visible scaffolding than usual. Annotation is adaptive. A student who normally needs two marks may sensibly need five on an unusually complex passage.

The key is that increased marking should follow structural need, not anxiety.

Structural difficulty versus emotional difficulty

Structural difficulty is observable: several similar entities, time shifts, distributed evidence, multiple qualifications, competing causal explanations. Emotional difficulty is the feeling that the passage is intimidating, technical or unfamiliar.

Annotation should respond mainly to structural difficulty. Marking every unfamiliar word because the topic feels hard can make the page more threatening without protecting any high-value relationship.

The no-mark threshold

A mark should be rejected when:

• the information is obvious and unlikely to be lost;
• the student can already return to the region instantly;
• the mark duplicates another stronger cue;
• the mark reflects uncertainty but does not identify what is uncertain;
• the mark gives visual emphasis to a speculative interpretation;
• the notation requires more decoding than rereading would;
• the mark is made only because the student always makes one in that position.

The must-mark threshold

No annotation category is universally mandatory, but some risks can become personally high-value. A student with repeated quantity omissions may temporarily treat explicit quantity words as must-mark triggers. A student with recurrent flashback confusion may mark time shifts until chronology stabilises.

Personal must-mark rules should be temporary and evidence-based. Once the error no longer recurs, the rule should be tested for retirement.

Annotation phases across one passage

Annotation can happen at different moments, and the correct kind of mark changes with the phase.

• First-read phase: structural anchors, time shifts, central contrast, provisional uncertainty.
• Question-reading phase: quantity, actor, time, relation, source and condition boundaries.
• Evidence-search phase: question-number anchors, evidence corridors, discriminating clue marks.
• Answer-construction phase: evidence ownership reminders or slot completion cues where needed.
• Verification phase: targeted return paths to evidence or question contract, not new broad annotation.
• Model-update phase: supersede stale marks visibly and locally.

Confusing the phases creates cost. Writing Q-numbers during first reading can prematurely predict questions. Adding new passage summaries during final verification can reopen already-settled structure.

First-read annotation should remain question-agnostic

Before the student knows which details will be tested, first-read marks should capture durable structure rather than imagined answers. Turning point, central contrast, time shift, important qualification and unresolved causal relationship can remain useful across many possible questions.

Predicting exact questions too early can bias what gets marked and what gets ignored.

Question-reading annotation should be contract-specific

Once the question appears, annotation can become highly specific. “Two”, “after”, “according to”, “how did X change”, or the named actor may receive a tiny visible cue because those words control the answer obligation.

Question marks should be cheap because they are local and temporary. The student does not need a full code; one circle or underline may be enough.

Evidence-search annotation should follow ownership

During search, mark the evidence corridor only after the relation to the question is understood. A keyword match is not yet ownership. Question-number annotation is strongest when it follows proof rather than anticipation.

Construction-phase annotation should be rare

Once evidence and proposition are secure, the student should usually write. Additional passage marking at this stage can indicate that proof was not truly complete. A small slot tick or evidence link can help students with known construction failures, but should fade with mastery.

Verification-phase annotation should not expand the page

Verification should use existing question and evidence anchors. Creating fresh global marks during checking is often a sign that the student is reopening the passage model instead of testing a defined risk.

Model-update annotation should overwrite priority, not merely add information

When later evidence changes the model, the visual hierarchy should make the new interpretation dominant. Cross out, amend or supersede the old mark. Do not leave two equally salient incompatible annotations and expect working memory to remember which one is current.

Anchor half-life

An annotation has a useful half-life: the period during which it continues to save retrieval or prevent error. Some marks remain useful for the whole passage set. Others matter only for one question.

A paragraph-role label may retain value across ten questions. A circled “two” matters only until the current answer is complete. Understanding half-life helps prevent temporary marks from continuing to occupy attention.

Long-half-life marks

• turning point;
• central paragraph role;
• time-shift boundary;
• key model update;
• major qualification controlling several later answers.

Short-half-life marks

• question quantity;
• one local actor reminder;
• question-number evidence link after answer closure;
• temporary uncertainty on one phrase;
• slot ticks used during construction.

Short-half-life marks can remain physically visible, but should become mentally inactive once their job is complete.

Re-entry latency

Re-entry latency is the time required to return to a passage region and restore enough context to continue reasoning. Good annotation should reduce this latency.

It is not enough for the student to reach the correct paragraph. The mark should help restore the relevant relationship: why the paragraph matters, what changed there, or which question it supports.

Location latency versus meaning latency

A mark can reduce location latency but not meaning latency. A star may lead the student to P4 instantly, yet the student still must reread P4 to remember whether it contains the cause, correction or result.

A functional label such as “correction” reduces both location and meaning latency.

Annotation quality can be measured by re-entry, not density

During practice, compare how long the student needs to answer “Where was the turning point?”, “Where is the qualification?”, or “Where is Q7’s evidence?” after a short delay. The best system is the one that restores the right meaning quickly with the lowest marking cost.

Annotation and changed answers

Annotations can make changed-answer verification much cheaper. If a changed answer is linked to a clear evidence corridor, the student can recheck the relevant passage relation rather than reread broadly.

But a Q-number should not create an obligation to revisit the answer. PCEP-011’s change threshold still governs whether review is necessary.

Annotation and final-review priority

Marks can support final review when they point to genuine high-risk answers: changed answer, distributed evidence, unresolved question, multi-part slot, strong inference. They should not determine review order simply because they are visually prominent.

The final-review queue belongs to verification. Annotation supplies routes, not priorities.

Annotation and passage cache safety

A passage mark can become part of the student’s cache. Safe cached marks represent explicit structure: P4 correction, P6 outcome, actor identity, time boundary. Unsafe cached marks represent unverified interpretation: “jealous”, “angry”, “failed completely”.

When annotation is used as external cache, certainty labels matter. The page should not make inference look like fact.

The cache-safety rule

Facts can be marked as facts. Inferences should stay visibly provisional until the evidence warrants commitment.

Annotation and pronoun ambiguity

Pronoun-reference instruction belongs to the specialist canonical page. Annotation’s job is narrower: when reference ambiguity threatens the passage model, a tiny entity cue can externalise the resolved ownership so the student does not solve the same pronoun repeatedly.

For specialist instruction, route to PSLE English Learning Guide: Pronoun Reference.

Annotation and multi-part scope

Multi-part scope instruction remains with its specialist owner. PCEP-012 asks how one or two small visible cues can protect that scope while attention moves into the passage.

For specialist scope work, route to PSLE English Learning Guide: Comprehension Multi-part Question Scope.

Annotation and inference evidence

Inference mechanics remain with the specialist owner. Annotation can preserve candidate clues, counterevidence and uncertainty without deciding the inference prematurely.

For specialist inference work, route to PSLE English Learning Guide: Comprehension Inference and Evidence.

Annotation and literal detail

Literal-detail instruction remains separate. Annotation can create a direct return anchor when the student repeatedly loses the local evidence location or time frame.

For specialist work, route to PSLE English Learning Guide: Comprehension Literal Detail.

Annotation and paraphrase

Paraphrase mechanics belong to the specialist owner. Annotation can protect semantic invariants by marking condition, time, quantity or certainty words before the student rewrites evidence into an answer.

For specialist paraphrase work, route to PSLE English Learning Guide: Comprehension Open-Ended Paraphrase.

Annotation and question reading

The specialist owner teaches how to interpret the question. PCEP-012 adds a performance interface: which part of that interpretation should be externalised because it is likely to disappear during evidence search?

For specialist question-reading work, route to PSLE English Learning Guide: Read the Question.

When annotation reveals a skill that needs practice

Annotation helps you preserve useful evidence, find it again, update your interpretation and compress notes. If a mark reveals difficulty with inference, reference, answer scope or paraphrasing, use the relevant linked guide for focused practice before returning to annotation.

Original annotation case 121: mark threshold correctly rejects obvious fact

One protagonist appears throughout a simple narrative. Maren chooses not to circle the name. Actor ownership never becomes uncertain. The absence of a mark is a correct annotation decision.

Original annotation case 122: mark threshold correctly captures exceptional risk

Four same-gender students enter one paragraph. Iona lightly labels two initials because later pronouns would otherwise be expensive to resolve repeatedly.

Original annotation case 123: emotional difficulty lowers threshold incorrectly

A passage about environmental chemistry feels unfamiliar. The student circles eleven technical nouns despite understanding the paragraph roles. Annotation density reflects anxiety rather than structural need.

Original annotation case 124: structural difficulty justifies extra marks

A passage contains flashback, present action and anticipated future event. Leonie uses three time anchors, more than her usual system, because event order genuinely needs protection.

Original annotation case 125: first-read mark remains question-agnostic

The student marks P4 as turning point but does not predict which question will use it. Later, both a change and cause question route through the same region.

Original annotation case 126: first-read annotation predicts too much

The student writes “Q likely asks confidence” beside a behaviour clue before seeing questions. When the actual item asks cause, the prediction biases attention.

Original annotation case 127: question-phase mark protects quantity

“Give another reason” receives one circle. During search, the student remembers the first reason is already known and deliberately seeks a distinct second one.

Original annotation case 128: question-phase mark protects source

“According to the survey” is boxed. The student prunes a later organiser opinion immediately.

Original annotation case 129: evidence-phase mark follows ownership

Only after confirming that P5 provides the effectiveness evidence does the student add “Q8”. The index is stable and useful.

Original annotation case 130: evidence-phase mark precedes ownership

A Q-number is written beside the first keyword match. Later proof lies elsewhere, but the visual anchor keeps pulling the student back to the wrong sentence.

Original annotation case 131: construction-phase tick prevents omission

Maren has two reasons secured. Two tiny ticks beside the question are crossed as each appears in writing. The scaffold prevents one known construction failure.

Original annotation case 132: construction-phase marking becomes planning essay

The student writes both reasons in full beside the question before composing the answer. The plan duplicates the answer and consumes more time than it saves.

Original annotation case 133: verification uses existing route

A changed answer has Q7 indexed beside its evidence corridor. Leonie checks only that corridor and the marked qualifier, then releases.

Original annotation case 134: verification adds new clutter

During checking, the student starts circling new details throughout the passage. Verification has become a new reading pass instead of risk testing.

Original annotation case 135: long-half-life mark pays repeatedly

“P4 turn” supports four later returns. Its value persists across the entire question set.

Original annotation case 136: short-half-life mark remains mentally active

A circled “two” from Q3 keeps drawing attention during later review even though Q3 is closed. The student has not learned mental retirement.

Original annotation case 137: re-entry location fast, meaning slow

A star takes the student to P5 instantly, but she still rereads all of P5 to remember why it matters. Location latency fell; meaning latency did not.

Original annotation case 138: re-entry location and meaning both fast

“P5 result” immediately restores both where and what. The student rereads only the exact outcome line.

Original annotation case 139: changed answer uses annotation safely

Q6 is changed after a source qualifier is noticed. The passage index helps verify the new answer without triggering a full reread.

Original annotation case 140: changed answer treats mark as authority

The student keeps a wrong answer because Q6 was already written beside a sentence. The annotation is mistaken for proof rather than a route marker.

Original annotation case 141: final-review route, not priority

Q8 has a useful passage index, but Q9 is the only flagged uncertainty. Final review goes to Q9 first. Annotation helps navigation without hijacking priority.

Original annotation case 142: visual prominence hijacks priority

A heavily boxed Q4 evidence zone is reviewed first despite Q9 being incomplete. Final-review order follows ink rather than risk.

Original annotation case 143: safe cache uses explicit structure

“P6 outcome” becomes a stable external cache. Later questions reuse the region without reinterpreting it.

Original annotation case 144: unsafe cache stores motive as fact

“jealous” is written in the margin after one clue. Later questions use the word as if the passage explicitly stated it.

Original annotation case 145: specialist handoff triggered correctly

The student marks pronoun chains but still resolves them wrongly. Annotation is not the first weak link. Training routes to specialist reference instruction.

Original annotation case 146: annotation is enough, specialist reteaching unnecessary

The student resolves pronouns accurately but forgets the actor after several questions. A tiny entity cue fixes the performance problem; pronoun instruction itself is already stable.

Original annotation case 147: multi-part skill stable, memory fragile

Maren understands multi-part scope perfectly but loses one slot during long evidence search. One quantity mark is sufficient support.

Original annotation case 148: multi-part skill itself weak

The student circles “two”, yet cannot identify what counts as two distinct reasons. Annotation cannot replace specialist scope instruction.

Original annotation case 149: inference uncertainty externalised safely

Iona marks “?” beside a motive and later replaces it with a supported interpretation after discriminating evidence appears.

Original annotation case 150: inference weakness hidden by marks

The student underlines all relevant clues but still consistently draws conclusions beyond the evidence ceiling. More marking will not repair inference calibration.

Original annotation case 151: literal-detail return anchor

A timeline mark allows the student to recover the exact event immediately after a time-qualified question.

Original annotation case 152: literal-detail instruction still needed

The student reaches the right sentence quickly but selects the wrong detail because evidence function is misunderstood. Annotation route is correct; comprehension micro-skill is not.

Original annotation case 153: paraphrase invariant protected

“most” is underlined in the evidence sentence. During paraphrase the student preserves “most” rather than writing “all”.

Original annotation case 154: paraphrase drift despite mark

The condition is visibly marked, but the student still changes “hesitated” to “refused”. Specialist paraphrase fidelity remains the real repair target.

Original annotation case 155: question-reading mark solves performance failure

The student knows how to answer change questions but forgets the word “later” under pressure. Marking the time boundary restores the stable skill.

Original annotation case 156: question-reading weakness survives mark

The student circles “why” but still cannot distinguish cause from purpose. The annotation exposes but does not repair the conceptual weakness.

Original annotation case 157: annotation category retired after migration

Maren now automatically notices quantity during question reading. Circling numbers no longer changes performance and is removed.

Original annotation case 158: category reactivated for stress test

During one unusually dense mock, late slot omissions return. Quantity marks are reintroduced temporarily while the execution weakness is stabilised.

Original annotation case 159: high-value page remains visually sparse

A six-paragraph passage contains only five marks: one time shift, one turning point, one qualification, one outcome and one Q-number corridor. Every mark is reused.

Original annotation case 160: dense page has low functional value

The passage contains twenty-three underlines and circles. Only two are used later. Annotation activity is high; annotation return is low.

Original annotation case 161: first-read anchor improves three questions

A central “turn” mark supports later cause, change and evidence questions. One question-agnostic structural cue has broad value.

Original annotation case 162: first-read predicted answer contaminates three questions

The student writes “fear” beside a character’s hesitation. Three later questions inherit the unsupported motive.

Original annotation case 163: annotation budget expands rationally

A passage with multiple time shifts receives two extra chronology marks. Re-entry and sequence accuracy improve enough to justify the cost.

Original annotation case 164: annotation budget expands emotionally

A technical-looking topic receives twice as many marks even though structure is simple. No search or accuracy benefit appears.

Original annotation case 165: long-half-life mark dominates appropriately

The one central model update remains visually salient because it controls several downstream answers.

Original annotation case 166: short-half-life mark dominates inappropriately

A heavily circled “two” from Q2 remains more visually striking than the later passage turning point. Temporary marks are consuming long-term salience.

Original annotation case 167: re-entry latency measured

Without annotation, Leonie needs twenty seconds to recover the outcome paragraph after interruption. With one “O” mark, she needs four. The cue has measurable return.

Original annotation case 168: mark saves location but not reasoning

A star cuts return time to P4, but the student still cannot remember whether P4 is cause or consequence. The cue needs better information scent.

Original annotation case 169: final-review annotation supports changed answer

A changed Q9 answer has one evidence corridor indexed. Verification is quick and evidence-led.

Original annotation case 170: final-review follows stale Q-number

Q9’s evidence changed after a model update, but the original Q-number was never moved. Review returns to outdated proof.

Original annotation case 171: cache fact safe across questions

“P6 outcome” remains a reliable structural cue for several items.

Original annotation case 172: cache inference unsafe across questions

“angry” remains in the margin after one tentative inference and is reused as fact later.

Original annotation case 173: specialist handoff prevents overengineering

A student keeps missing pronoun reference despite elaborate entity marks. The tutor stops adding annotation complexity and repairs the underlying reference skill directly.

Original annotation case 174: performance scaffold avoids unnecessary reteaching

Another student resolves pronouns correctly but forgets who acted after six questions. One entity anchor solves the performance memory issue without reteaching pronouns.

Original annotation case 175: mock-paper annotation remains affordable

The student’s first-read marks take little time, are reused throughout the set and remain interpretable late in the paper. Annotation survives full simulation.

Original annotation case 176: mock-paper annotation collapses under pressure

A complex training code is abandoned halfway through the timed set because it is too expensive. The system has not crossed the execution threshold.

Original annotation case 177: final-revision system shrinks

In the final weeks, four annotation categories are reduced to two: question boundary and passage-return anchor. Accuracy remains stable and speed improves.

Original annotation case 178: final-revision system expands unnecessarily

A new colour code is introduced days before the examination despite no recurring annotation problem. Decoding cost rises and the established routine destabilises.

Original annotation case 179: exam-day system degrades gracefully

Late in the paper, Leonie stops margin labels but keeps one-risk question marks and one evidence corridor. The highest-value functions survive time pressure.

Original annotation case 180: complete annotation architecture

The student marks only when future retrieval or boundary risk justifies it; distinguishes first-read structure from question-specific evidence; uses stable, high-information cues; keeps candidate inference visibly provisional; updates stale marks; indexes evidence only after ownership is known; allows temporary marks to become mentally inactive; uses annotation to reduce re-entry latency; hands genuine micro-skill weaknesses to specialist training; fades scaffolds as competence becomes internal; and preserves only the smallest examination-day interface that still protects marks.

Annotation drill 41: mark/no-mark threshold

Present twenty candidate phrases from a passage. Students choose which deserve marks and justify each decision by future use, expected error or retrieval cost.

Drill 42: structural versus emotional difficulty

Compare an unfamiliar but structurally simple passage with a familiar but structurally complex passage. Annotation density should follow structure rather than comfort.

Drill 43: phase discipline

Separate first-read, question, evidence-search and verification phases. Allow only annotation appropriate to each phase. Identify marks made too early or too late.

Drill 44: anchor half-life

Classify marks as long-half-life or short-half-life. Practise mentally clearing temporary marks after answer closure.

Drill 45: re-entry latency

Measure the time needed to recover meaning from marked and unmarked passage regions after an interruption. Optimise for both location and meaning latency.

Drill 46: changed-answer route

Index one changed answer to its evidence corridor, then verify it under a short time limit. Compare with no index.

Drill 47: final-review priority separation

Give a heavily annotated page and a separate risk list. Students must choose review order from risk, using annotation only for navigation.

Drill 48: cache-safety audit

Classify margin notes as explicit fact, structural role, provisional inference or unsupported interpretation. Unsafe cache entries are relabelled or removed.

Drill 49: specialist-handoff diagnosis

Give several errors and ask whether annotation can solve the performance problem or whether the underlying micro-skill needs direct repair.

Drill 50: pronoun-anchor economy

Use a multi-character passage. Add the smallest entity marks that prevent repeated pronoun resolution without circling every name.

Drill 51: flashback anchor economy

Use a passage with non-linear chronology. Students design the smallest visible time structure that preserves event order.

Drill 52: comparison-index economy

Mark one shared comparison dimension in the question and one matched evidence phrase per side. Avoid marking unrelated differences.

Drill 53: condition/exception economy

Use passages with general rules and exceptions. Students mark only the boundary words that alter the target case.

Drill 54: claim/qualification economy

Mark a claim and the smallest qualification controlling certainty. Compare with full-paragraph annotation.

Drill 55: criterion/outcome indexing

Place matching marks at goal and outcome regions for evaluation questions. Confirm that the marks do not themselves imply success.

Drill 56: multi-part question interface

Use one question-side quantity cue and minimal passage-side evidence anchors. Fade slot ticks once construction becomes reliable.

Drill 57: scaffold-retirement gate

Remove one annotation category after three stable sets. Restore it only if a matching error recurs under changed conditions.

Drill 58: mock-paper annotation audit

After a mock, count marks created, marks reused, false anchors, stale marks, global rereads and time spent annotating. Identify the single category with the best and worst return.

Drill 59: final-revision compression

Reduce the system to its smallest proven examination form. Compare two timed sets before and after compression.

Drill 60: full annotation transfer test

Use an unseen passage in a different genre and topic, under realistic timing and one deliberate disruption. The system passes only if marks stay sparse, accurate, interpretable, updateable and genuinely useful for re-entry.

Advanced twelve-week path for strong readers

Weeks 1–2 audit overannotation and identify which marks have real return. Weeks 3–4 train no-mark decisions and one-risk question cues. Weeks 5–6 focus on provisional inference marking, counterevidence and visible model updates. Weeks 7–8 reduce re-entry latency with paragraph-role and evidence-corridor cues. Weeks 9–10 pressure-test the sparse system under unfamiliar informational and narrative texts. Weeks 11–12 retire redundant categories and confirm that low annotation density does not increase error.

The strong-reader goal is not minimal ink for its own sake. It is maximum structural control per mark.

Advanced twelve-week path for developing readers

Weeks 1–2 use explicit paragraph-role labels and entity anchors. Weeks 3–4 stabilise chronology and state change. Weeks 5–6 add question-contract marks and evidence corridors. Weeks 7–8 introduce provisional uncertainty and model-update marks. Weeks 9–10 compress notes and reduce symbol categories. Weeks 11–12 test whether internal structure survives scaffold removal under timed unseen work.

The developing-reader goal is not to remove supports quickly. It is to remove them only after the underlying representation has moved inside.

Maren’s long annotation trajectory

Maren begins by underlining too much of the question and passage. Her first repair is one-risk question marking: quantity, time or actor only. She then learns dual indexing, using one question-side boundary and one passage-side route. As scope becomes automatic, question marks fade. Her final system is exceptionally sparse because the earlier visible scaffolds have migrated into task reading.

Iona’s long annotation trajectory

Iona begins with sophisticated but overcommitted notes. She learns candidate-versus-confirmed status, uncertainty marks and visible counterevidence. Later she practices model version control so a first interpretation can be superseded cleanly. Her final system protects only exceptional uncertainty, qualification and distributed evidence.

Leonie’s long annotation trajectory

Leonie begins with accurate but elaborate marking. She measures marking time, re-entry latency and marks actually reused. Full-sentence notes shrink to function labels, then to two or three stable anchors. Late-paper compression is trained deliberately so the system degrades gracefully instead of collapsing. Her final annotation exists mainly to reduce recovery and repeated search.

Tutor annotation dashboard

• mark count by passage;
• annotation time;
• marks reused;
• false anchors;
• stale marks after updates;
• global versus targeted rereads;
• re-entry latency;
• question-scope errors;
• annotation categories ready to retire;
• one active annotation target.

The dashboard should lead to one teaching decision. If it becomes a large reporting system, it has reproduced the very annotation problem it is meant to solve.

Monthly annotation audit

Once a month, compare recent unseen passages. Is total marking falling while route accuracy remains stable? Are marks reused? Are stale marks being corrected? Does one genre trigger saturation? Are any categories no longer catching errors? The audit decides what to preserve, compress, retire or temporarily reactivate.

Mock-paper annotation audit

A mock tests whether annotation survives real sequencing and pressure. Measure not only marking time but its downstream effects. Did anchors reduce rereading? Did one difficult question cause the page to become denser? Did final review follow risk or visible ink? Did annotations remain legible and current after several model updates?

Annotation retirement gate

• the matching error stays absent across varied passages;
• the student recognises the risk automatically;
• removing the mark does not increase rereading or mistakes;
• timed performance remains stable;
• delayed retest remains stable.

If the gate is passed, the category moves to maintenance or disappears from the visible routine.

Annotation maintenance

Maintenance does not mean continuing every old mark. It means keeping the ability to reintroduce a mark when a structural risk returns. The student owns a flexible toolkit rather than a fixed code.

Final examination compression

By examination day, the system should feel almost invisible:

Mark only what you may lose, need to revisit, or must keep provisional. Let structure live mostly in the mind. Let the page carry only the risks that deserve external memory.

For many strong students, that may mean only a handful of marks. For a developing reader on a structurally dense passage, it may mean more. The number is not the standard. Functional return is.

The full annotation-performance architecture

Understand first → assess retrieval or boundary risk → mark only if external memory has value → preserve uncertainty honestly → index evidence after ownership → use marks to re-enter, not to replace thinking → update stale marks → mentally clear temporary marks → retire scaffolds when the skill becomes internal.

Final canonical boundary

PCEP-012 does not claim ownership over question-reading, inference, pronoun reference, paraphrase, passage mapping, evidence selection or verification. It owns the external interface connecting those mechanisms to the printed comprehension environment under time.

That interface deserves its own standard because bad annotation can create false certainty, clutter, premature commitment and wasted time, while good annotation can preserve scope, reduce re-entry latency, protect model updates and make the whole comprehension system cheaper.

Final deep return to Maren, Iona and Leonie

Maren’s page is no longer covered in evidence she already understands. She marks the boundary most likely to disappear while she searches, then lets the contract do its work.

Iona’s page no longer turns every intriguing clue into a conclusion. Her marks preserve the difference between fact, candidate interpretation and confirmed evidence, so later information can update the model without resistance.

Leonie’s marks no longer prove how carefully she read. They prove their value by making the next return cheaper. When a mark no longer saves time or prevents error, she stops making it.

Closing deep principle: annotation is successful when the page becomes quieter and the thinking becomes clearer

Good annotation does not compete with the passage.

It does not turn uncertainty into false certainty.

It does not force every question into a colour or symbol system.

It leaves behind a small number of reliable visual cues that help the student remember structure, preserve boundaries, return to evidence and update meaning.

When those cues have finished their jobs, the student moves on.