Why Reading Comprehension Marks Are Lost | Diagnose the Failure Point

51. Diagnose blank answers by asking why nothing became writable

A blank answer can arise from several different failures. The learner may not understand the command, may not find evidence, may see too many interpretations, may run out of time, or may believe an imperfect answer is worse than no answer. Do not code every blank as “did not know”. Reconstruct the final known state. Was an evidence corridor identified? Was one interpretation leading? Did the learner know part of the answer but refuse to write it? Blank-answer diagnosis should identify the earliest missing component that prevented knowledge from becoming a scorable response.

52. Diagnose incomplete multi-part answers

A multi-part question may lose marks because the learner does not notice the second demand, answers two parts with one idea, or runs out of time after completing the first. Mark the response against the question’s visible components before judging the quality of prose. If the learner understood both parts orally but omitted one in writing, the first failure is answer-contract control. If the learner never recognised the second part, question parsing failed earlier. If time ended before the second part, execution becomes the first cause.

53. Diagnose over-answering

Over-answering is not merely a style issue. It can consume time, dilute scope and introduce contradictions. Compare what the question required with what each sentence contributed. If sentence three adds no new scoring value, ask why it was written. Some learners over-answer because they do not trust the first point; others because they confuse explanation with quantity. The repair may be mark-value awareness, response contracts or confidence calibration. “Write less” is too vague unless the mechanism producing excess is identified.

54. Diagnose under-answering

A concise answer can be precise or incomplete. Under-answering occurs when required components are missing: one side of a comparison, evidence without effect, a cause without explanation, one of two requested details. Ask whether the learner knew the missing component before writing. If yes, expression or completion checking failed. If no, question interpretation or reasoning failed upstream. This distinction prevents teachers from treating every short answer as a language-production weakness.

55. Diagnose answer-length drift after a difficult question

Answer length can change after a confidence shock. A learner who normally writes one precise sentence may start producing defensive paragraphs after one uncertain inference. If later answers become longer without higher mark value, diagnose the first question that changed the behaviour. The visible timing problem may be caused by confidence rather than writing speed. SETC-SEC-COMP-070 owns the recovery routine; 080 uses the answer-length trace as evidence that recovery failed to restore normal execution.

56. Diagnose source-attribution errors under fatigue

A learner may attribute sources accurately early but blend them late in a long section. This pattern differs from a stable source-ownership misconception. Compare early and late paired-source questions. If untimed and early performance is strong but late accuracy drops, the first cause may be fatigue or working-memory overload. The repair should simplify late-paper source tracking and increase realistic-duration practice rather than reteach the concept of attribution from the beginning.

57. Diagnose question-order errors

A learner can lose marks because the chosen order increases switching cost, destroys locality or leaves a high-value task too late. Review the route, not only the answers. Did the learner jump to “easy” items across several pages and repeatedly reload context? Did a high-mark comparison lose its protected window? If so, the first failure is sequencing. SETC-SEC-COMP-040 owns the routing mechanics; 080 identifies whether route choice was the causal origin of the later score loss.

58. Diagnose skipping too late

If a question consumes excessive time before being left, the first diagnostic question is whether progress had already stalled. Look for repeated scans, unchanged alternatives or rewriting without new evidence. If the learner could have detected stalled state earlier, the repair belongs to ceiling recognition and recovery. If meaningful progress continued until the leave point, the overrun may have been a reasonable investment. Timing alone does not decide whether the skip was late.

59. Diagnose skipping too early

A learner may flag questions before performing the ordinary routine. The result is a return queue filled with items that were never truly blocked. Examine whether command classification, local search and one reasoning attempt occurred before the skip. If not, the first cause is avoidance or an overly low skip threshold. The repair is minimum engagement, not more reserve at the end. Efficient recovery depends on selective queue entry.

60. Diagnose return-queue failures

A skipped question can still cost marks because it is forgotten, returned to in the wrong order or restarted from zero. Inspect the flag, breadcrumb and return timing. If the learner remembers the item but has to rebuild every step, state preservation failed. If a nearly complete high-value item waits behind low-value uncertain items, return prioritisation failed. If the learner never revisits despite available reserve, the planned return window failed. Different queue failures need different repairs.

61. Diagnose time-debt cascades

When the final page is rushed, reconstruct when the debt began and how the learner reacted. One overrun may be harmless if later work remains normal. The cascade begins when the learner underreads the next passage, skips too aggressively, overchecks or loses answer scope. Diagnose both the original debt and the first maladaptive repayment behaviour. Repairing only the original slow question may not prevent the same collapse if the response to debt remains unstable.

62. Diagnose interface-navigation errors

In digital assessments, some lost marks originate in navigation rather than English. The learner may scroll away from evidence, reopen the wrong question, forget a flag or misread a split-screen layout. Reconstruct the mechanical path. If the conceptual answer is correct once the right screen is restored, the first repair belongs to interface fluency. Mock practice should reproduce the actual medium closely enough that mechanical weaknesses become visible before high-stakes use.

63. Diagnose annotation overload

Heavy annotation can feel organised while adding little retrieval value. Compare what was marked with what the learner later used. If six colours, multiple symbols and extensive underlining do not reduce search time or improve evidence accuracy, the annotation system is producing cost without return. The first repair is subtraction: keep paragraph jobs, turns or source markers that actually aid navigation. Annotation should function as an index, not a transcript of attention.

64. Diagnose under-annotation or missing landmarks

The opposite pattern occurs when the learner has no reusable structure and repeatedly searches from the top. The issue is not that every passage needs visible notes; expert readers may hold structure mentally. The diagnostic test is retrieval performance. If evidence corridors are hard to predict and broad rereads are common, the passage map is insufficiently externalised or internalised. A minimal paragraph-function map may be the smallest useful repair.

65. Diagnose mechanical answer-transfer errors

A correct response can lose marks if entered in the wrong place, left on the question booklet when a separate sheet is required, or shifted one row in a multiple-choice grid. These are not comprehension errors. They belong to the complete examination workflow. Review whether transfer was budgeted and whether the learner had a final mechanical check. Training should include the real response medium so mechanical steps become part of the performance system rather than an afterthought.

66. Diagnose attraction to multiple-choice distractors

Distractors often contain a true detail attached to the wrong question job, an exaggerated degree, a reversed cause, or an inference one step beyond the evidence. Ask why the chosen option felt attractive. If the learner cannot identify the exact evidence supporting it, selection may have been familiarity-based. Compare the final two options on the precise dimension where they differ. The first failure is often scope or relation, not broad comprehension.

67. Diagnose elimination without positive evidence

A learner may eliminate three options for weak reasons and select the remaining one without proving it. This can work by luck. Diagnosis should ask for positive support for the chosen answer, not only objections to alternatives. If the final option cannot be connected to evidence, the process is unstable even when correct. Multiple-choice strategy should end with evidence confirmation, especially when distractors differ by subtle scope or certainty.

68. Diagnose premature closure on global-purpose questions

A learner may decide the writer’s purpose from the opening and then force later evidence into that frame. When a final purpose answer is wrong, inspect when the interpretation was first fixed. If the learner never updated after concessions or a changed conclusion, the first cause is model rigidity. The repair is to treat early purpose as provisional until major turns and final stance have been processed.

69. Diagnose local-versus-global scale mixing

A critical sentence does not make the whole article hostile; a hopeful ending does not erase an earlier cautious paragraph. If the learner uses local evidence for a whole-text claim, label the textual scale of question and evidence. If both are mismatched, the first repair is scale control. Many tone, purpose and stance errors are actually failures to keep word, paragraph, section and whole-text levels separate.

70. Diagnose one-sided change-over-time answers

A question asking how attitude changes requires at least an earlier and later state. If the learner gives only the final state, determine whether the first state was missed during retrieval or omitted during expression. The distinction matters. A retrieval failure calls for distributed-evidence training; an expression failure calls for answer-structure control. Change questions expose whether the learner understands evidence geometry across time.

71. Diagnose title and headline framing errors

Headlines can exaggerate while still pointing to a real trend; titles can develop a second meaning over the text. If the learner treats framing as literal fact or ignores its later reinterpretation, ask when the title was last reconsidered. The first failure may be whole-text integration rather than vocabulary. Good diagnosis distinguishes factual proposition from rhetorical framing and early expectation from final significance.

72. Diagnose graph-variable errors

Wrong graph interpretations often begin before trend reading. The learner may confuse axes, units, groups or time periods. Ask for the variable pair in words before discussing increase or decrease. If the variables are wrong, no later evaluation should be diagnosed yet. Visual comprehension begins with measurement identity. A dramatic-looking slope is irrelevant if the learner has attached it to the wrong group.

73. Diagnose caption and image overreading

Photographs invite emotional inference that may exceed what the task permits. If the learner calls a subject sad, frightened or ashamed based only on posture, inspect caption and surrounding text. The first failure may be projection. Conversely, if the caption supplies a crucial place or date and the learner ignores it, multimodal integration failed. Image evidence should be scoped like prose evidence: what does it show, what does the caption add, and what remains unknowable?

74. Diagnose statistics: relevance versus sufficiency

A statistic can be relevant without proving the entire claim. If a learner treats a local survey as proof of a national conclusion, the issue may be representativeness or scale. If the learner dismisses useful data because it is imperfect, the issue may be all-or-nothing evaluation. Diagnosis should identify what proposition the statistic actually supports and what additional evidence would be required for the broader claim.

75. Diagnose reliability and credibility errors

Students sometimes treat expert titles as automatic proof or treat bias as automatic invalidation. Ask which criterion is relevant: expertise, method, sample, incentive, corroboration, direct access or transparency. If the learner uses a vague label such as “reliable because expert”, evaluation reasoning is underdeveloped. The repair is criterion-based judgement, not memorising a list of trustworthy source types.

76. Diagnose representativeness errors

One anecdote can show existence but not prevalence. A local survey can show one population but not every population. If the learner generalises beyond the sample, scope failed. If she does not notice what sample was used, evidence-quality reading failed earlier. The distinction determines whether training should focus on source reading or evaluation logic.

77. Diagnose assumption-cost errors

Some interpretations are possible only because the learner supplies missing backstory. Ask what must be true for the answer to work. If several invisible assumptions are required while a simpler reading explains the same evidence, inference inflation is likely. Training can compare two readings explicitly and count which facts come from the text versus imagination. The best-supported reading usually carries lower assumption cost.

78. Diagnose certainty overstatement

A learner may move from “suggests” to “proves” during writing. Compare evidence type with final verb. Direct statement can usually support declarative language; inference may require “suggests”, “appears” or another calibrated form. If the reasoning is good but certainty is too strong, do not reteach the inference. Repair answer calibration. This is a common high-level loss among learners who equate confident language with strong analysis.

79. Diagnose unnecessary hedging of direct facts

Some learners write “may have”, “perhaps” and “possibly” even when the passage explicitly states the event. This can make a correct answer sound less precise. Ask whether the hesitation comes from poor memory, fear of commitment or confusion between stated and inferred information. The repair is different from overstatement. Calibration must work in both directions.

80. Diagnose mixed stance flattened into one label

A writer can support an aim but criticise implementation. A narrator can feel affection and resentment simultaneously. If the learner chooses one extreme label and ignores the second dimension, ask whether the evidence actually conflicts or operates on different axes. The first repair is dimensional separation. Nuance is not the same as indecision; it is a structured account of multiple supported relations.

81. Diagnose irony errors

Irony often fails when the learner reads literal wording without the contextual mismatch, or detects sarcasm but cannot explain what creates it. Ask for the literal layer and the surrounding reality separately. If the learner can state both but not connect them, reasoning failed. If she never notices the mismatch, evidence or passage-model reading failed earlier. The repair depends on which layer is missing.

82. Diagnose symbolism projection

Generic associations—rain means sadness, light means hope—can override local evidence. Ask whether the object recurs, appears at structural turns or changes meaning. If the symbolic claim depends only on a memorised association, it is projection. The repair is pattern-based inference. Symbolism should emerge from recurrence and placement rather than from a prelearned dictionary of images.

83. Diagnose literary ambiguity overgeneration

Strong students sometimes generate too many plausible readings and lose time or fail to commit. Review whether every alternative has evidence. If only one or two do, the problem is possibility inflation. If two remain genuinely supported, certainty calibration and decision thresholds become the repair. SETC-SEC-COMP-060 owns the full ambiguity system; 080 identifies whether overgeneration caused the mark loss or timing damage.

84. Diagnose background-knowledge intrusion

A learner who knows the topic may answer from outside knowledge rather than the passage. Ask which part of the response is explicitly supported by the text. If the unsupported component comes from subject expertise, the first failure is source discipline. Background knowledge can help interpret vocabulary and context, but it should not replace the evidence when the question is passage-bound.

85. Diagnose missing local definitions

Technical passages often define unfamiliar terms through apposition, examples or restatement. If the learner guesses from prior knowledge while the passage provides its own definition, the first failure is local reading. Train recognition of definitional structures before prescribing broad vocabulary study. Sometimes the answer was already inside the text.

86. Diagnose syntax overload

Dense sentences can produce wrong answers because the learner never stabilises the main clause, condition or contrast. Ask the learner to strip the sentence to its core proposition and then restore modifiers. If meaning becomes correct after decomposition, syntax processing is the repair point. This differs from a conceptual inference weakness and should be trained with sentence-level parsing before full-passage pressure returns.

87. Diagnose cohesion and connective errors

Missed words such as “however”, “therefore”, “for example” and “even so” can reverse the relationship between ideas. If the learner’s passage map treats a concession as a conclusion or an example as a new claim, connective reading failed. This can contaminate several questions. Repairing relationship words may yield more benefit than practising tone or purpose separately.

88. Diagnose concession errors

A concession admits something without necessarily surrendering the main position. If the learner interprets “admittedly, the plan is costly” as opposition even though “nevertheless” leads to support, the first cause is discourse structure. Train concession + surviving claim as a pair. The learner should be able to say what the writer grants and what the writer still concludes.

89. Diagnose writer-versus-expert confusion

Articles may quote voices they later challenge. If the learner uses an expert’s certainty as the writer’s certainty, inspect the reporting frame. Verbs such as “claims”, “argues”, “admits” and the sentences that follow matter. The first repair is ownership plus stance relation, not broad evaluation. Source identity should be stable before asking what the writer believes.

90. Diagnose evidence-versus-explanation confusion

A learner may copy a quotation and assume it explains itself. Or she may give an interpretation with no textual support. Diagnose which side is missing. Evidence without explanation needs a reasoning bridge; explanation without evidence needs retrieval. Treating both as “weak answer” hides opposite repair needs.

91. Diagnose high-cost correct answers

A correct answer that takes excessive time can create downstream mark loss. Record the longest-question times even when correct. Ask what consumed the minutes: broad search, multiple alternative readings, answer rewriting or repeated checking. The first cause may be process inefficiency rather than knowledge. High-cost correctness is a leading indicator of future timing problems and deserves repair before it becomes visible in scores.

92. Diagnose lucky correct answers

If the learner cannot explain why the chosen answer is supported, correct scoring may be luck. During selected review, ask for evidence after marking is hidden. If no defensible route can be reconstructed, classify the item as process-unstable rather than mastered. This prevents inflated confidence and helps practice target reliability rather than one-off outcomes.

93. Diagnose high-confidence wrong answers

These answers are diagnostically valuable because the learner is unlikely to self-correct without feedback. Look for stable misconceptions, source misattribution, overlearned generic formulas or premature closure. The repair should include a contrast case showing why the original rule fails. Merely telling the learner to “check more” may worsen timing without changing the underlying belief.

94. Diagnose low-confidence correct answers

A learner who repeatedly doubts correct answers may waste time checking and may change them later. Compare evidence quality with confidence. If strong support consistently produces low confidence, calibration is the repair target. Training can require the learner to state why the evidence is sufficient and then stop. Confidence should become more aligned with process quality over time.

95. Diagnose recurring errors versus one-off errors

One error is a clue; a repeated mechanism is evidence. Maintain a short ledger of first failure points across fresh texts. If “missed concession” appears three times in different topics, it deserves targeted repair. If “misread one unusual caption” never recurs, a large intervention may be unnecessary. Recurrence protects the learner from overfitting training to one paper.

96. Diagnose clusters instead of isolated questions

When several errors appear in one passage, ask whether they share an upstream cause. A missed final stance can affect purpose, tone, comparison and evaluation. If one repair explains the cluster, it has higher leverage than four separate drills. Error clusters are often where the first-failure-point model produces the greatest training efficiency.

97. Diagnose the cascade chain

A useful post-mortem can write the cascade explicitly: hard inference → four-minute overrun → confidence shock → overchecking → final summary rushed. Only the first event is the local comprehension problem; the rest are execution consequences. Repair may need both levels: improve inference efficiency and install containment. Writing the chain prevents later symptoms from being mistaken for independent weaknesses.

98. Prioritise root causes by leverage

Not every identified weakness deserves equal practice time. High-leverage causes affect many questions, recur across texts and cost meaningful marks or time. A missed passage turn may deserve priority over one rare vocabulary slip. A chronic answer-scope problem may deserve more work than an isolated punctuation misunderstanding. Diagnosis becomes actionable only when repair order is chosen.

99. Choose the smallest repair that can falsify the diagnosis

If the suspected cause is keyword dependence, do not assign five full papers. Use a short paraphrase-distance drill on fresh text. If the learner succeeds, the diagnosis may be wrong or already repaired. If the problem persists, expand training. Small tests are efficient because they distinguish mechanism from coincidence before large practice volume is added.

100. Retest on fresh text before declaring the repair successful

Correcting the original question proves that the learner can understand feedback, not that the skill has transferred. Use an unseen passage with different surface vocabulary, topic and layout but the same underlying mechanism. If the learner now recognises the concession, preserves the quantifier or builds the comparison axis independently, the repair is becoming portable. Fresh transfer is the final diagnostic check.

101. Practice Lab: four wrong answers begin with one missed turn

Maren gets four questions wrong in one argumentative passage: tone, purpose, final stance and evaluation. It would be easy to assign four separate drills. The passage map tells a different story. She recorded the opening claim and evidence but missed the paragraph beginning “Even so”, where the writer narrows the recommendation.

All four downstream answers inherit the wrong whole-text model. The first repair is recognising and updating at structural turns. A fresh passage with a concession and narrowed conclusion tests whether that mechanism transfers.

One upstream repair can explain a cluster of marks.

102. Practice Lab: evidence correct, inference too strong

Iona identifies repeated hesitation, a broken sentence and avoidance of eye contact. She answers that the character feels guilty about lying. The evidence supports discomfort but not guilt or lying.

The passage map and evidence search are both sound. The first wrong state appears at the inference bridge, where an unsupported motive is added. Her repair is smallest-necessary inference: uncomfortable or anxious unless a clue specifically supports guilt.

Do not reteach evidence search when the evidence was already right.

103. Practice Lab: the final answer is vague because the question was misread

Maren writes that an anecdote “makes the article more interesting”. The question asks how the anecdote changes the reader’s understanding of the problem. When asked orally what the anecdote does, she can explain that it turns abstract cost into a personal consequence.

Her conceptual reading is stronger than the written answer suggests. The first failure was question-job control: she answered a generic purpose question rather than the specific effect-on-understanding task.

The repair is command paraphrase before evidence search.

104. Practice Lab: the right reasoning begins from the wrong paragraph

Iona explains a character’s refusal logically using evidence from the refusal scene. The passage earlier revealed that the character had promised not to repeat a previous mistake. The examiner expects that earlier motive.

The reasoning is coherent, but the evidence corridor is wrong. The diagnosis should stop before inference. Her repair is to distinguish event location from cause location and search backward when the question asks why.

Logical reasoning cannot rescue the wrong evidence set.

105. Practice Lab: the right paragraph contains the wrong sentence

Leonie reaches the correct paragraph for a question about why public trust falls. She selects a sentence describing service delays because it looks relevant. The next sentence says officials repeatedly promised the delays were temporary and were later proved wrong.

The region is correct. Evidence quality fails because the selected sentence identifies the problem but not the reason trust changes. The repair is relation matching inside the corridor.

Search diagnosis must distinguish region from proposition.

106. Practice Lab: “many” becomes “everyone”

Maren finds the exact sentence: “Many participants improved when weekly support continued.” Her answer says the programme helps everyone.

Nothing upstream is wrong. The quantifier disappears during paraphrase. Her repair is proposition fidelity: preserve actor, degree, condition and relation in a compact evidence note before drafting.

A small word can be the first failure point.

107. Practice Lab: cautious becomes opposed

Iona correctly notices repeated “may”, “appears” and “the evidence remains limited”. She labels the writer opposed to the proposal. The conclusion actually recommends another trial.

The evidence cluster is good; the tone label is too strong. Certainty and stance calibration are the first repair, not passage modelling.

Near-neighbour labels can be the final point where a good process loses the mark.

108. Practice Lab: a correct answer takes five minutes

Leonie earns the mark on a one-mark vocabulary question but spends five minutes choosing between two close meanings. Her total score later falls because the last page is rushed.

The vocabulary mark is locally correct and globally expensive. The diagnostic target is not vocabulary knowledge alone; it is decision threshold and value-aware timing.

Correct answers can still reveal the mechanism that loses later marks.

109. Practice Lab: a wrong answer is completed in twenty seconds

Maren answers a global-purpose question almost immediately after reading the opening. She is confident and wrong. Later paragraphs qualify the initial claim.

The fast wrong answer is diagnostic of premature closure. The repair is not “spend more time on every question”. It is keep global interpretations provisional until major passage turns and conclusion are integrated.

Speed is evidence when it reveals where reasoning stopped too early.

110. Practice Lab: the learner changes a correct answer

Iona’s first inference is supported and concise. During checking she imagines a more sophisticated motive and rewrites the answer. The new motive requires facts the passage never supplies.

The first answer process was sound. The mark is lost during verification. Her repair is error-triggered checking: change only when a specific contradiction, missing component or source error is identified.

The cause can occur after the correct answer already exists.

111. Practice Lab: a blank answer was almost complete mentally

Leonie leaves a two-mark inference blank. In review she immediately states one supported clue and a reasonable interpretation. She says she did not write because she was “not completely sure”.

The first cause is not lack of comprehension. It is certainty threshold and conversion of partial knowledge into scorable form. Training should practise writing defensible qualified answers before certainty feels perfect.

A blank can hide usable knowledge.

112. Practice Lab: a two-part answer contains only the first part

Maren identifies how a phrase creates urgency but does not explain why that matters to the writer’s purpose, even though the question explicitly asks both.

When the question is paraphrased before answering, she notices both demands. The first failure is command parsing, not language expression.

Missing components should be traced to when they first disappeared.

113. Practice Lab: the answer is too long because the learner mistrusts one point

Iona gives one correct reason, then adds two weaker possibilities “just in case”. One of them contradicts the first and the response loses precision.

Evidence search and reasoning produced a sufficient answer. The first failure is confidence-based overcollection after sufficiency. Her repair is a stopping criterion tied to mark value and answer contract.

More content is not always more security.

114. Practice Lab: a short answer is incomplete, not elegant

Leonie writes “to create tension” for a question asking how a sentence creates tension. The label is plausible but no mechanism connects the language to the effect.

When asked what the sentence does, she can explain that the short clauses compress action. The oral reasoning exists. Expression and answer-form completion are the repair points.

Conciseness is not the same as completeness.

115. Practice Lab: paired sources blend only late in the section

Iona answers early source-specific questions accurately but confuses Source A and B in the final comparison. Untimed she separates them easily.

The pattern points to fatigue and working-memory load rather than a conceptual source-ownership gap. Her repair is late-paper source labelling under realistic duration.

Same symptom, different mechanism depending on when it appears.

116. Practice Lab: easy-first creates repeated context reload

Maren answers questions in perceived difficulty order. She earns many easy marks but spends excessive time flipping among passage regions and forgets one skipped comparison.

The answer errors look varied, yet the route shows a common cause: fragmented question order. A printed-order control set reduces time and forgotten work.

Route reconstruction can explain seemingly unrelated losses.

117. Practice Lab: one late skip destroys the final task

Leonie spends six minutes on a low-value ambiguous item before finally skipping. The final summary receives half its normal working time.

The summary errors are downstream. Practising more summaries would not address why the time disappeared. The first repair is earlier block detection and containment.

Diagnose the cause before the page where the marks are visibly lost.

118. Practice Lab: six flags reveal avoidance

Iona finishes the first pass with six unresolved items. Review shows several were skipped before a serious local search or reasoning attempt.

The return problem begins at queue entry. Her repair is minimum engagement, not a larger reserve or more elaborate return strategy.

The earliest bad decision often occurs long before the final queue becomes visible.

119. Practice Lab: a return starts from zero

Maren flags a comparison with only a star. On return she rereads both sources and recreates the same two candidate axes.

The first skip decision was reasonable; state preservation was not. Her repair is a breadcrumb containing source positions and missing axis.

Diagnosis can clear one recovery stage while identifying the next one.

120. Practice Lab: the final page is rushed because checking consumed the reserve

Leonie reaches a checkpoint slightly ahead and uses the surplus to reread every completed answer. She later rushes the last high-value task.

The visible error is timing. The first cause is reserve misuse through non-risk-based checking. Her repair is to carry surplus forward until first completion.

Time diagnosis should trace where reserve was spent, not only where it ran out.

121. Practice Lab: the learner scrolls to the wrong paragraph

A digital interface returns to the top after each question. Iona repeatedly relocates evidence and once answers from a similar phrase in the wrong paragraph.

Untimed reasoning is fine. The first cause is mechanical navigation. Practice with paragraph landmarks and the real interface reduces the error without reteaching inference.

Not every wrong answer is an English weakness.

122. Practice Lab: six-colour annotation looks organised but costs four minutes

Maren highlights names, techniques, tone, evidence, questions and vocabulary in different colours. Later she uses only paragraph functions and one contrast marker.

The final timing loss begins during annotation. A stripped-down system preserves retrieval accuracy and saves minutes.

Diagnostic review should ask which markings actually changed later decisions.

123. Practice Lab: no landmarks create three broad rereads

Iona reads accurately but makes no mental or visible map. Each question triggers a fresh scan. She finishes late despite few conceptual errors.

The first repair is not faster reading. It is reusable passage representation. A compact paragraph-job map reduces later search cost.

Timing symptoms can originate in representation architecture.

124. Practice Lab: the correct answer is entered on the wrong line

Leonie reasons correctly but transfers two responses one row down on the answer sheet.

The first cause is mechanical workflow, not comprehension. Her next mock includes answer transfer inside the section budget and a row-alignment check.

Diagnosis should protect the learner from solving the wrong problem with more reading practice.

125. Practice Lab: a distractor contains a true detail

A multiple-choice option repeats a real event but answers “what happened” instead of “why it happened”. Iona chooses it because it looks textually familiar.

The first cause is question-job mismatch, not factual memory. Her repair is to compare options against the command relation, not against topic overlap.

True details can still be wrong answers.

126. Practice Lab: elimination leaves one unsupported option

Maren rejects three choices but cannot connect the remaining option to evidence. It happens to be correct.

The score should not hide process instability. Her repair is a final positive-evidence check after elimination.

A correct result can still fail the transfer test.

127. Practice Lab: headline drama overrides chart scale

A headline says “Participation Soars”. The chart shows an increase from 48 to 53 per cent. Leonie describes a dramatic surge.

The first cause is framing dominating measurement. Her repair begins at axes and magnitude before headline interpretation.

Multimodal diagnosis asks which source should carry the burden of each claim.

128. Practice Lab: a photograph is treated as citywide evidence

Iona uses one crowded photograph to prove that public transport use has risen across the city.

The image proves one instance. A chart or wider data source is needed for scale. The first failure is evidential scope, not image description.

Evidence of existence is not evidence of prevalence.

129. Practice Lab: the expert title becomes automatic proof

Maren says a claim is reliable because “a professor said it”. The passage gives no field of expertise and the expert comments outside the study’s measured area.

The first cause is criterion-free credibility judgement. Her repair is to ask what makes this source relevant to this claim.

Authority labels should not replace evaluation.

130. Practice Lab: one anecdote becomes proof of a trend

Iona uses a compelling personal story as proof that most families experience the same problem.

The anecdote is relevant but not representative. The first repair is claim-scale control: one case can illustrate existence without establishing frequency.

Evaluation errors often begin with evidence-to-claim scale mismatch.

131. Practice Lab: the dramatic interpretation requires hidden facts

Maren interprets a character’s silence as evidence of a secret betrayal. The passage supports discomfort but provides no betrayal clue.

When assumptions are listed, the dramatic reading needs several invisible facts. A simpler embarrassment reading explains the observed behaviour.

Assumption cost reveals inference inflation.

132. Practice Lab: direct fact is hedged into uncertainty

The passage explicitly says the meeting was cancelled. Iona writes that it “may have been cancelled”.

The first cause is certainty undercalibration, possibly driven by general exam anxiety. The repair is classify evidence as stated or inferred before choosing answer verbs.

Precision sometimes requires stronger language, not weaker language.

133. Practice Lab: mixed support is flattened into opposition

A writer supports a policy’s goal but criticises implementation. Leonie labels the writer opposed because the criticism is vivid.

The first failure is dimensional collapse. Her repair separates principle, implementation and scale before selecting a stance label.

Strong negative local language need not define the whole position.

134. Practice Lab: irony is missed because the literal word is positive

The narrator calls a failed plan “brilliant”. Maren records positive tone.

When literal wording is compared with surrounding failure, the mismatch becomes clear. The first cause is failure to integrate local wording with contextual reality.

Irony diagnosis needs two layers of evidence.

135. Practice Lab: rain becomes sadness by habit

Iona sees rain and writes “symbolises sadness” even though the characters celebrate outside.

The first cause is imported generic symbolism. Her repair requires recurrence, placement and contextual function before symbolic claims are allowed.

Memorised associations can be a source of error, not only knowledge.

136. Practice Lab: three plausible readings consume five minutes

Maren can imagine jealousy, anxiety and pride. Only anxiety and pride have textual support; anxiety explains more clues.

The first cause is possibility overgeneration and failure to rank by evidence. SETC-SEC-COMP-060 supplies the repair; the diagnostic signal is time spent after the evidence hierarchy was already sufficient.

Ambiguity becomes a performance problem when it lacks a stopping rule.

137. Practice Lab: topic expertise overrides the source

Iona knows a policy debate and writes the usual criticism, but the passage takes a narrower position.

Her first repair is source discipline: mark which clauses come from the passage and which come from memory. Outside knowledge can support understanding but cannot replace passage evidence.

Knowledge becomes harmful when it controls the answer instead of the text.

138. Practice Lab: one complex sentence creates three wrong answers

A dense sentence contains the main claim, a condition and a concession. Maren reads the condition as the main claim. Three later answers inherit the error.

The first repair is sentence decomposition, not three question-family drills. Once the proposition is parsed correctly, the downstream cluster disappears on fresh material.

Sentence-level processing can be a high-leverage root cause.

139. Practice Lab: one missed “however” changes the whole article

Iona maps the article as continuous support because she skims past a “however” introducing limitations. Purpose, tone and final stance all become wrong.

The error ledger should record “missed discourse turn”, not three separate comprehension weaknesses.

Small connective words can own large downstream consequences.

140. Practice Lab: the learner finally repairs the right thing

Across three papers, Leonie loses different marks: tone, comparison and evaluation. The common trace is that she repeatedly drops conditions such as “only if” and “unless” when paraphrasing evidence.

Instead of practising all three question families, she completes a short condition-preservation drill on fresh passages, then retests the original families. Accuracy improves across all of them.

This is the purpose of first-failure-point diagnosis: fewer repairs, chosen better.

---

Technical Appendix A: The comprehension first-failure taxonomy

A diagnostic system becomes useful when every lost mark can be placed at the earliest stage that plausibly caused it. The categories below are ordered from upstream representation to downstream execution. The order matters: later defects should not be labelled as root causes until earlier stages have been cleared.

• Passage model: chronology, viewpoint, paragraph jobs, structural turns, source identity, final stance.
• Question job: command, requested relation, textual scale, number of components, mark-value expectations.
• Evidence region: wrong paragraph, wrong source, wrong time frame, global search where local search was required.
• Evidence quality: topical but insufficient sentence, missing condition, wrong side of contrast, weak representativeness, omitted second state.
• Reasoning: inference inflation, inference underreach, causal confusion, mismatched comparison axis, poor evaluation criterion.
• Scope: degree, certainty, quantifier, time frame, source ownership, local versus global claim width.
• Expression: unclear relation, missing component, pronoun ambiguity, distorted paraphrase, wording that changes the idea.
• Checking: missed blank, ungrounded answer change, failure to verify known risk.
• Timing and routing: slow search, answer inflation, poor question order, late skip, reserve misuse.
• Recovery: cascade after one block, forgotten queue, poor state preservation, failure to restore ordinary execution.
• Mechanical workflow: digital navigation, answer transfer, interface mismatch, response-location error.

A single wrong answer can contain several categories. The diagnostic task is not to list all of them. It is to identify the first unstable stage that made later failures more likely.

Technical Appendix B: Question-family failure map

Question families produce characteristic failure patterns. These patterns are hypotheses, not automatic labels.

• Direct detail: wrong evidence region, reference error, paraphrase distance, scope distortion.
• Inference: incomplete evidence set, assumption inflation, certainty overstatement, failure to distinguish narrator belief from textual fact.
• Vocabulary in context: dictionary-memory dominance, missed grammar/contrast/restatement, overlong uncertainty loops.
• Language effect: technique naming without effect, generic effect labels, local evidence detached from surrounding function.
• Tone: near-neighbour degree errors, local/global scale mix, failure to track change, modality ignored.
• Purpose: whole-text generic purpose used for local function, example detached from parent claim, premature closure.
• Structure: feature list without before/after movement, missed turn, paragraph jobs not represented.
• Comparison: source blending, mismatched axis, one-sided evidence, asymmetric degree.
• Evaluation: criterion not named, relevance confused with sufficiency, authority substituted for credibility reasoning.
• Summary: category error, duplication, lost proposition in paraphrase, word-limit damage, selection and writing mixed.
• Multimodal: wrong variable, visual salience overriding relevance, caption ignored, scale inferred from one image.

Technical Appendix C: The first-cause interview

After marking, a teacher can reconstruct one important error using a short interview. The questions should follow the original attempt rather than the correct solution.

1. What did you think the passage section was doing?
2. What did you think the question wanted?
3. Where did you look first, and why?
4. Which exact evidence did you use?
5. What did you think that evidence meant?
6. How certain were you?
7. What did you decide the answer needed to contain?
8. Did you change the answer later?
9. How long did the question take relative to normal?
10. Did this question change how you handled the next questions?

The first answer in the sequence that reveals an incorrect state is the leading diagnosis. Retest before treating it as stable.

Technical Appendix D: Marked-script reconstruction

A marked script can be reconstructed as a timeline rather than a score sheet. For each diagnostic item, note the original answer, any crossed-out answer, annotations, evidence marks, confidence, approximate time and whether the item was revisited. Then write the attempt as a sequence: “read command as purpose → searched paragraph six → selected anecdote → answered global persuasion → checked but did not change.” The first mismatch with the task becomes visible.

This process is especially useful when students say, “I knew it but wrote it wrong.” Sometimes that is true. Sometimes the marked trace shows the misunderstanding began much earlier. The reconstruction gives teacher and learner a shared object of analysis instead of relying on memory.

Technical Appendix E: A comprehension error ledger that stays small

The ledger should not become a database of every wrong answer. Use one row per meaningful mechanism. Recommended fields are: date, passage type, question family, mark loss, first failure point, evidence for diagnosis, smallest repair, fresh transfer test, result and whether the mechanism recurred later.

Example: “Argument / tone / 1 mark / missed concession in passage map / learner map omitted ‘however’ paragraph / drill: identify concession + surviving claim in 5 short passages / fresh test next week.” This is much more actionable than “Tone weak”.

Archive mechanisms that stop recurring. The active ledger should remain short enough that training priorities are visible.

Technical Appendix F: Confidence × correctness matrix

Confidence adds useful diagnostic information when recorded before marking.

• Correct + high confidence: likely stable, but check cost if unusually slow.
• Correct + low confidence: possible calibration problem, repeated checking risk or lucky process.
• Wrong + high confidence: priority signal for misconception, premature closure, source error or overlearned formula.
• Wrong + low confidence: expected uncertainty; diagnose whether the learner lacked evidence, reasoning or execution control.

Over time, good calibration means confidence becomes more aligned with evidence quality. The objective is not to make students feel confident about everything; it is to make confidence informative.

Technical Appendix G: Time × correctness matrix

• Correct + normal time: healthy baseline.
• Correct + excessive time: inefficient but currently successful; inspect search, ambiguity and checking loops.
• Wrong + very fast: premature closure, guessed option, shallow command reading or untested assumption.
• Wrong + excessive time: stalled process, poor recovery or unresolved conceptual gap.

This matrix prevents score-only review. High-cost correctness and fast wrongness can both predict future performance instability.

Technical Appendix H: Repair-priority matrix

Prioritise a mechanism when it scores highly on four dimensions: recurrence, mark/time cost, downstream leverage and transfer breadth. A missed passage turn may affect many question families and recur across arguments. A rare unusual word may affect one question once.

A practical order is: high recurrence + high leverage first; high cost but rare second if the consequence is severe; frequent low-cost slips next if they accumulate; isolated low-impact errors last. The matrix keeps dramatic but rare mistakes from displacing more important recurring mechanisms.

Technical Appendix I: One-week diagnosis-to-repair cycle

Day 1: run a fresh mixed comprehension set and reconstruct two or three meaningful errors. Day 2: isolate the highest-leverage suspected mechanism with a short untimed drill. Day 3: vary surface wording while holding the mechanism constant. Day 4: recombine the mechanism into a short passage. Day 5: add moderate timing. Day 6: use an unseen transfer passage. Day 7: review whether the mechanism recurred and either archive, refine or continue the repair.

One week should normally target one major mechanism and perhaps one secondary mechanism. Too many simultaneous repairs make attribution difficult.

Technical Appendix J: Six-week repair programme

Week 1: Baseline diagnosis. Build a small ledger and identify one or two recurring first failure points. Week 2: Upstream repair. Work passage modelling, question interpretation or evidence retrieval if the diagnosis lies there. Week 3: Reasoning and scope. Target inference, comparison, evaluation, certainty or answer contracts. Week 4: Recombination. Put repaired mechanisms back into full passages with interleaved question types. Week 5: Execution. Add timing, question order, recovery and realistic medium. Week 6: Transfer and simulation. Use unseen material and full comprehension sections; diagnose only the mechanisms that still recur.

The programme should get simpler as the ledger shrinks. If the list of active problems grows every week, diagnosis may be too symptom-based.

Technical Appendix K: 30-day comprehension repair programme

Days 1–5: collect trustworthy evidence

Use fresh material. Record original answers, confidence on selected items, longest-question time and visible process traces. Build the first-failure ledger only after reconstructing attempts.

Days 6–10: isolate the first major mechanism

Use short drills that make the suspected failure easy to observe: concession mapping, paraphrase-distance search, condition preservation, comparison-axis selection or inference-size control. Retest with fresh surface forms.

Days 11–15: introduce competing demands

Interleave the repaired mechanism with other question types so success does not depend on knowing what skill is being tested. Keep timing moderate.

Days 16–20: add execution

Use full passages and section checkpoints. Observe whether the repaired skill survives fatigue, ambiguity, question-order decisions and one controlled recovery event.

Days 21–25: transfer across text families

Move among narrative, argument, informational, paired and multimodal texts. A mechanism is stronger when it survives changes in topic, genre and layout.

Days 26–29: full-section confirmation

Run realistic comprehension sections. Use the ledger after each one, but add a new repair only when the mechanism is recurring and consequential.

Day 30: simplify the active repair list

Archive mechanisms that have transferred reliably. Keep only the two or three highest-value active weaknesses for the next cycle. A good repair programme produces a shorter operating list over time.

Technical Appendix L: Teacher review protocol

1. Choose errors with high leverage or unusual process cost, not every lost mark.
2. Hide the model answer initially where possible.
3. Ask the learner to reconstruct passage state and question job.
4. Inspect original evidence marks and annotations.
5. Clear upstream stages one by one.
6. Name the first wrong state in neutral operational language.
7. Choose the smallest repair that targets that state.
8. Test the repair on a fresh example during the same or next session.
9. Record the mechanism only if it is meaningful enough to track.
10. Retire the mechanism after repeated successful transfer.

Teacher diagnosis should produce fewer labels and more testable hypotheses. “Careless” is not a useful first cause. “Dropped the condition during paraphrase on three passages” is.

Technical Appendix M: Self-study protocol

After marking, select no more than three items. For each, write: “I thought the question wanted…”, “I looked at…”, “My evidence was…”, “My reasoning was…”, “The first stage that went wrong was…”, “My next fresh test will be…”. Keep the notes brief.

If the learner cannot identify the first stage confidently, compare the attempt with a teacher, model explanation or trusted marking guidance. Do not force a diagnosis merely to fill a ledger. Uncertain diagnoses should remain hypotheses until retested.

Technical Appendix N: Parent support protocol

Parents do not need to mark literary interpretations. They can help the learner make post-mortems concrete. Useful questions include: “Where did the answer first go off track?” “Did you misunderstand the text, the question, the evidence or the timing?” “Is this the same mistake as last week?” “What fresh question will test the repair?”

Avoid turning every lost mark into more worksheets. The diagnostic model is designed to reduce unnecessary volume by identifying what actually needs practice.

Technical Appendix O: Full post-mortem — one narrative passage

Imagine a narrative in which a narrator waits to meet an estranged friend. The learner loses four marks: one motive inference, one tone question, one title question and one comparison of early versus late attitude. At first these look like four weaknesses.

Reconstruction begins with the passage map. The learner recorded “narrator worried → friend arrives angry → argument → reconciliation”. The text, however, never establishes that the friend is angry. That is the narrator’s assumption. The first wrong state is viewpoint: narrator belief was treated as textual fact.

The motive inference then uses the supposed anger to explain silence. The tone answer calls the meeting hostile throughout. The title interpretation treats “The Return” as recovery from conflict rather than return to a misunderstood relationship. The early-versus-late attitude answer exaggerates the shift because the early state is already misrepresented.

The repair is not four separate drills. It is a viewpoint-control drill: observable event versus narrator interpretation, followed by a fresh limited-narration passage. On retest, the learner marks who knows what and distinguishes reported perception from external fact. Several question families improve together.

This is high-leverage diagnosis: a single first wrong state explains the error cluster and produces a compact repair.

Technical Appendix P: Full post-mortem — one argumentative passage

Imagine an argument about extending a public programme. The learner loses marks on final stance, evidence evaluation and comparison with a second source. She also runs out of time.

The passage map is accurate until paragraph six, where the writer says, “Although the pilot is promising, its small sample cannot justify nationwide expansion yet.” The learner underlines “pilot is promising” but not the limiting clause. Evidence selection therefore preserves only one side of the concession.

From that incomplete proposition she concludes the writer supports expansion. The evaluation answer calls the pilot strong proof. The comparison with Source B becomes falsely polarised because Source B also wants more evidence. The learner then spends extra time trying to reconcile two sources that are actually closer than her model suggests.

The first repair is proposition fidelity inside concessive sentences, especially condition and degree. A short drill on “although / but / nevertheless / only if” structures is followed by fresh source comparison. The timing issue is treated partly as downstream because the false source contrast increased reasoning cost.

Technical Appendix Q: Score-plateau diagnosis

A plateau does not necessarily mean the learner needs harder passages or more papers. Review the active error ledger. If the same first failure point continues recurring, the repair has not transferred and more full papers may only repeat it. If the error mechanisms are changing and shrinking, the plateau may reflect tougher material rather than stagnant skill. If untimed accuracy rises while timed scores stay flat, execution is now the bottleneck.

Use leading indicators: fewer broad rereads, better passage maps, lower longest-question time, improved confidence calibration, smaller return queues and fewer scope distortions. Score is important, but it can lag behind mechanism improvement. A plateau becomes actionable only after its internal composition is understood.

Technical Appendix R: From diagnosis to the next lane spoke

Diagnosis decides what to practise. The next lane article, SETC-SEC-COMP-090, owns the comprehension-specific transfer architecture that follows: isolate the diagnosed mechanism, vary the surface, recombine it with other skills and test on unseen texts. Diagnosis without transfer practice becomes explanation without improvement. Practice without diagnosis becomes volume without targeting.

This separation preserves canonical ownership. 080 answers “what failed first?” 090 answers “how do we repair it so the skill survives new passages?”

141. How SETC-SEC-COMP-080 connects to the lane

This page owns comprehension-specific first-failure-point diagnosis. It routes upstream repairs to the relevant lane spokes: passage modelling to SETC-SEC-COMP-020, timing to 030, question order to 040, evidence search to 050, ambiguity to 060, and recovery to 070.

The broader cross-paper error-analysis owner remains English Exam Error Analysis. Return to SETC-SEC-COMP-000 for the complete Secondary English comprehension performance system.

142. Final principle: repair where the failure began

A marked answer shows the endpoint of an attempt. Good diagnosis reconstructs the path. The learner may have lost the mark in one sentence, but the repair may belong several stages earlier—in the passage map, command, evidence corridor, certainty threshold or recovery decision.

Iona traces the evidence chain. Maren separates representation from expression. Leonie distinguishes skill from execution. Their shared purpose is to prevent practice from becoming indiscriminate volume.

DO NOT REPAIR THE PLACE WHERE THE ERROR BECAME VISIBLE UNTIL YOU HAVE FOUND THE PLACE WHERE IT FIRST BECAME INEVITABLE.