Reading Faster Without Losing Comprehension | The Exam Speed–Accuracy Problem

51. Build a speed profile from operations, not identity labels

“I am a slow reader” is too broad to train. A useful profile separates orientation, syntax repair, vocabulary resolution, passage mapping, evidence search, inference, answer planning, writing and checking. One operation may be slow while others are already efficient. The profile turns speed from a personality judgement into a list of observable costs.

52. Profile orientation speed

How long does the learner take to identify topic, voice, controlling question and passage direction? If orientation is slow because every introductory detail is treated as equally important, the repair is hierarchy. If orientation is fast but wrong because the learner assumes the writer’s stance from the title, the repair is provisional modelling. Orientation speed should create a useful initial frame without premature closure.

53. Profile syntax-repair speed

Record how often dense sentences cause the learner to stop and how long local repair takes. If one difficult sentence consumes several minutes, teach a compact decomposition routine: main proposition, relation, condition, qualifier. If repair becomes faster while later answers remain accurate, genuine processing speed has improved.

54. Profile vocabulary-resolution speed

Some learners stop at nearly every unfamiliar word. Others ignore important lexical uncertainty and answer wrongly. The profile should distinguish unnecessary pauses from necessary ones. Track whether the word controls a key claim or question. The speed goal is proportionate resolution: solve what matters, postpone what does not.

55. Profile discourse-integration speed

A reader may understand every sentence yet be slow because paragraph relations are rebuilt repeatedly. Ask the learner to state how each paragraph changes the model: adds evidence, concedes, contrasts, narrows, exemplifies, reframes. When these relations become automatic, passage reading and later retrieval both accelerate.

56. Profile evidence-recognition speed

This is not the same as evidence-search architecture in 050. Here the question is cost: once the learner reaches the right region, how quickly can she recognise which sentence actually answers the command? If every relevant-looking sentence receives equal attention, search closes slowly. Train relation matching: what proposition is required, and which sentence carries it?

57. Profile inference-closure speed

Some inference work is slow because evidence is hard to find. Other inference work is slow because the learner does not know when to stop reasoning. Record how long the learner continues after sufficient evidence appears. If extra time produces only additional unsupported possibilities, the bottleneck is closure rather than comprehension.

58. Profile answer-planning speed

Before writing, how long does the learner take to decide what the response needs? A long planning pause can reveal unclear answer contracts. Practise identifying logical slots from command and mark demand. Planning should become short because the structure is recognised, not because the learner writes before knowing what she is doing.

59. Profile writing speed separately from thinking speed

If the evidence and answer skeleton are ready but writing remains slow, inspect expression. Is the learner searching for impressive vocabulary, rewriting sentences before finishing them, or copying long quotations? If writing is fast once the decision state is stable, the earlier bottleneck was thinking. Separate the two before prescribing practice.

60. Profile verification speed

Verification can be cheap when it tests one defect or expensive when it becomes full re-analysis. Record how long checking takes and what it finds. If two minutes of verification repeatedly add no value, train narrower triggers. If a thirty-second scope check catches recurring errors, preserve it. Speed should remove low-yield verification, not all verification.

61. False speed appears when later repair rises

A reader may appear faster because the first pass is shorter, yet broad rereads, answer changes and scope errors increase. This is false speed. The learner has borrowed time from later stages. Any compression experiment should track downstream repair so hidden debt becomes visible.

62. False speed appears when answer completeness falls

Finishing earlier by omitting second components, explanations or source symmetry is not improvement. The speed metric must include scoring completeness. If answer time falls but multi-part marks disappear, the learner crossed the safe compression boundary.

63. False speed appears when scope becomes broader

Quantifiers, conditions and certainty are often the first things lost when a learner rushes. Monitor them deliberately during speed training. A passage read twenty per cent faster is not useful if “many” repeatedly becomes “all”. Scope fidelity is a sensitive control variable because the core topic can remain correct while the proposition becomes wrong.

64. False speed appears when source identity blurs

Paired texts are especially vulnerable. Faster reading may preserve general topic but lose who said what. If source swaps rise as time falls, source separation has not yet become automatic. Slow down that operation, strengthen the source map, and compress elsewhere.

65. False speed appears when the learner answers from familiarity

Fast recognition can be mistaken for fast comprehension. A familiar phrase or topic may trigger a remembered answer pattern. The learner then writes before checking whether the current passage uses the idea differently. Include familiar-surface contrast cases in training so speed remains text-controlled.

66. Safe speed comes from chunking meaningful units

As comprehension improves, the learner no longer processes every word as an isolated event. Common phrase patterns, clause relations and paragraph jobs become larger meaningful units. Training should support this through repeated exposure to syntax, connectives, discourse roles and question families. The goal is not skipping words; it is recognising larger units accurately enough that fewer decisions are needed.

67. Safe speed comes from prediction that remains revisable

A heading, opening claim or paragraph transition can create useful predictions about what comes next. Prediction reduces processing cost because the reader has an expected frame. But prediction must remain revisable. If new evidence conflicts, the model updates. Speed comes from prepared interpretation, not from forcing the text to confirm the first guess.

68. Safe speed comes from lexical automaticity

Common academic and examination vocabulary should not require repeated conscious decoding. Broader reading and vocabulary learning reduce the cost of recognising connectives, evaluative language, reporting verbs and abstract nouns. 120 does not own the vocabulary curriculum, but it treats lexical fluency as one reason the same passage can become easier to process over time.

69. Safe speed comes from syntactic familiarity

Repeated exposure to concessive clauses, relative clauses, conditionals and reporting structures makes them easier to parse. The learner begins to recognise the relationship before consciously naming every grammatical part. Syntax practice therefore contributes to speed when it reduces local repair without losing proposition accuracy.

70. Safe speed comes from discourse familiarity

An experienced reader recognises common movements: problem → causes → response, claim → evidence → concession → conclusion, event → reflection → decision. These patterns are not rigid templates, but they reduce orientation cost. The reader asks which pattern the text is using rather than treating every paragraph as structurally novel.

71. Do not chase headline “speed-reading” claims inside examination training

Any technique that promises dramatic acceleration should be tested against the actual task. Can the learner still identify passage movement, preserve conditions, locate evidence and answer unfamiliar questions? If not, the technique is irrelevant to comprehension examination performance. 120 rewards demonstrated lower total cost with stable accuracy, not impressive reading-rate numbers detached from scoring.

72. Skipping words is not a general solution

Some low-value detail can be read lightly, but indiscriminate omission is risky because small words can control logic. The safer principle is variable resolution: process structural and scoring-critical information carefully, move more quickly through supporting material whose role is already clear, and return locally when a question upgrades its importance.

73. Forced uniform pace is inefficient

A learner who tries to maintain the same speed through a simple anecdote and a dense concessive argument wastes capacity in one and risks errors in the other. Productive speed is elastic. Train acceleration and deceleration as parts of the same skill.

74. The learner should know where she is allowed to slow down

Speed anxiety often causes learners to rush the exact places that require precision. Build a protected list: major turns, conditions, negation, quoted-source ownership, final stance, comparison axis, dense causal chains. Slowing locally can improve total speed by preventing later repair. Leonie treats selective slowdown as an efficiency move, not a failure.

75. The learner should know where she is allowed to move lightly

Repeated examples, decorative background, predictable transitions and details whose function is already understood may not need deep first-pass analysis. The learner still reads them, but the resolution is lower. If a later question targets them, close reading can occur then. This preserves attention for information with higher expected structural value.

76. Speed should survive a change in passage length

Short-passage fluency may rely on memory rather than indexing. Increase length gradually so the learner must depend on passage structure. If speed collapses only when texts become longer, the missing skill may be representation and retrieval rather than raw reading rate.

77. Speed should survive a change in density

Use passages of similar length but different conceptual density. The learner should slow appropriately on denser text without losing the whole section. The goal is not identical time; it is efficient adaptation. Stable performance across density changes is stronger evidence than one fast average.

78. Speed should survive a change in genre

Narrative, argument and informational texts distribute difficulty differently. A mature reader carries abstract operations across them: chronology, relation, evidence, turn, conclusion. Genre variation tests whether speed comes from transferable reading control or from familiarity with one text family.

79. Speed should survive unfamiliar question wording

If a learner becomes slow when “What does this suggest?” changes to “What can be inferred?”, the question-family recognition is still cue-dependent. Vary surface wording so command classification becomes faster at the conceptual level. Recognition speed is part of answer speed.

80. Speed should survive mixed question types

Blocked drills can make a learner fast because the next operation is known in advance. Mixed sets remove that artificial cue. The learner must first recognise whether the task is literal, inferential, comparative, evaluative or structural. Examination speed requires operation selection as well as operation execution.

81. Build a compression ladder for passage modelling

Start from an accurate comfortable model. Then reduce annotation, shorten paragraph labels, remove redundant highlighting and gradually lower first-read time. At every step, test whether the learner can still state major movement, turn, voices and final position. Stop compression when one of those begins to fail.

82. Build a compression ladder for syntax repair

Begin with explicit clause marking on difficult sentences. Then reduce the visible support: underline main clause only, then perform the same decomposition mentally. Later embed one dense sentence in a normal passage and test whether the learner can repair locally without slowing everything around it. The goal is a faster repair routine, not avoiding repair.

83. Build a compression ladder for vocabulary decisions

Early practice can require full contextual analysis. Later, use a quick sequence: necessary or nonessential? grammar? contrast? local definition? best-supported meaning? The learner should learn when to stop. Speed improves as familiar context cues become automatic and low-value unknown words are postponed.

84. Build a compression ladder for inference

Give evidence first and require the smallest supported conclusion. Then add one competing interpretation. Next remove the supplied corridor. Finally interleave inference with other question types under moderate time. The learner should become faster because evidence sufficiency and assumption cost are recognised more quickly.

85. Build a compression ladder for comparison

Begin with two source summaries and ask only for the shared axis. Then require evidence. Next use full sources. Later vary the axis wording and source length. The speed gain comes from making axis selection and source separation automatic enough that rereading both texts broadly is unnecessary.

86. Build a compression ladder for answer construction

First allow answer skeletons on paper. Then reduce them to one or two keywords. Later, generate the structure mentally before writing. Compare time and completeness. If shorter planning creates missing components, restore support. The target is fast structure recognition followed by concise expression.

87. Build a compression ladder for checking

110 owns the checking system; 120 contributes the compression question. Which checks must remain even when the reserve shrinks? Blanks, missing components, source/scope risks and mechanics usually survive longer than stylistic review. Practise reducing checking depth while preserving detection of high-cost defects.

88. Use paired metrics: time plus passage-map quality

A faster passage read counts only if the resulting map remains usable. After fresh passages, score a few map elements: main movement, important turn, voices, final position. Compare that with first-read time. Improvement means time falls while map quality stays stable or rises.

89. Use paired metrics: time plus evidence accuracy

For local questions, record how long evidence selection takes and whether the selected proposition is sufficient. Faster selection with more wrong regions is false speed. Faster selection with the same evidence quality is a genuine gain.

90. Use paired metrics: time plus answer completeness

Answer construction should be judged against required components. Track time and whether every scoring job is present. A concise answer that remains complete is faster; a short answer missing the explanation is merely incomplete.

91. Use paired metrics: time plus scope fidelity

Quantifier, condition, certainty and time frame are excellent accuracy controls during speed training. They reveal whether the learner is compressing safely. If these details degrade first, the current pace is too fast for proposition-level precision.

92. Use paired metrics: time plus reread count

A faster first read with more rereads may not be faster at all. Track both. Healthy compression often reduces or stabilises rereads because the passage map becomes more efficient. If rereads rise, inspect what was lost in the first pass.

93. Use paired metrics: time plus answer-change rate

Speed that increases later corrections can create hidden cost. Record answers changed because the first pass was too shallow, not changes caused by genuinely new later evidence. If compression produces more repair, the system may have crossed its safe boundary.

94. A speed plateau is diagnostic information

If time stops improving while accuracy remains high, the learner may have reached the current efficient floor for that process. Further compression should not be automatic. Look for another bottleneck or build more fluency through broad reading and language development. 130 will own the broader score-plateau problem; here the point is that not every stable time is a failure.

95. Speed should become quieter as automaticity grows

Early training uses visible routines: label the turn, decompose the sentence, name the inference bridge, mark the comparison axis. Later, these decisions should require less conscious effort. The learner still performs them, but they no longer consume the same working memory. Speed growth often looks like fewer explicit steps, not more frantic execution.

96. Do not compress every operation to the same degree

Some operations have large waste margins; others are already near their safe minimum. Annotation may shrink dramatically. Negation processing may not. Answer wording may shorten. Complex comparison may still need substantial thinking. The speed programme should target where waste exists rather than impose uniform reduction.

97. Protect irreversible decisions more than reversible ones

A rough passage prediction can be revised later; a submitted answer under a no-return interface may not be. A preliminary tone hypothesis can stay provisional; an answer changed in the final seconds may replace a correct one. Where decisions are hard to reverse, accuracy deserves more protection. Speed should account for the cost of error as well as the time of action.

98. Build speed from broad reading outside test practice

Not every speed gain should come from timed drills. Regular reading across narrative, argument, science, history, journalism and other domains builds familiarity with vocabulary, syntax and discourse patterns. This lowers processing cost in unseen texts without teaching the learner to rush. Examination practice then teaches how to deploy that fluency under constraints.

99. Build speed from language control outside comprehension papers

Grammar, vocabulary and sentence-level reading practice can remove bottlenecks before they appear in full passages. If the learner routinely struggles with embedded clauses or reporting verbs, repeated comprehension mocks may be an inefficient way to repair them. Component fluency supports later whole-task speed.

100. Mature speed means selective precision at lower total cost

The mature learner does not try to make every moment faster. She makes the overall system cheaper. She moves lightly through predictable support, slows at structural risk, maps just enough to navigate, searches locally, closes inference when evidence is sufficient, writes to the scoring job and verifies only real risk. The result is not reckless speed. It is lower total processing cost with preserved meaning.

101. Practice Lab: Maren cuts the first read and creates three rereads

Maren forces herself to finish an argumentative passage two minutes faster. She reaches the questions early but has not registered the concession or narrowed conclusion. Three later questions send her back to the whole text. Total completion becomes slower. The repair is to restore structural reading while trimming only low-value annotation. Her first read becomes slightly longer again, but total passage cost falls.

102. Practice Lab: Iona slows only at one dense sentence

Iona reads the passage efficiently until one sentence contains a concession, condition and long relative clause. She rereads it four times. Training teaches local decomposition: main proposition first, then qualification. On fresh passages she now spends extra seconds only on the difficult sentence and returns immediately to normal pace. Local slowdown makes the whole passage faster.

103. Practice Lab: Leonie stops for every unfamiliar word

A science passage contains six unfamiliar terms. Leonie tries to resolve every one before continuing. Only two matter to the argument. Her new triage asks whether the word controls a key relation or question. Four terms become placeholders. The passage remains coherent because cause, contrast and conclusion are clear. Total reading time falls without loss of answer accuracy.

104. Practice Lab: a known word produces false fluency

Maren sees a familiar word and applies its everyday meaning immediately. The sentence becomes subtly incoherent, but she moves on quickly. A later vocabulary-in-context question exposes the error. Her repair is a fast contextual substitution whenever familiar meaning conflicts with grammar or surrounding logic. The extra second prevents a larger misread.

105. Practice Lab: a title prediction saves time until it becomes a trap

Iona uses the title to predict the writer will oppose a policy. Early paragraphs support that frame, so reading feels fast. A late concession changes the stance, but she fails to update. The initial speed gain becomes a wrong global answer. Training keeps title predictions provisional and explicitly checks major turns. Prediction remains useful because it is now revisable.

106. Practice Lab: narrative description is over-read

Leonie analyses every image in a long descriptive opening even though later questions focus on motive and change. The first-pass map only needed setting, viewpoint and emotional pressure. She practises reading descriptive detail lightly unless recurrence or a question upgrades it. Passage time drops while inference accuracy remains stable.

107. Practice Lab: argument paragraphs are treated as unrelated topics

Maren understands each paragraph individually but repeatedly asks, “Why is this here?” The text feels slow because every paragraph is reinterpreted from zero. She begins labelling discourse roles—claim, evidence, concession, counterargument, narrowed conclusion. Once paragraph function becomes predictable, integration speeds up and later purpose questions require less rereading.

108. Practice Lab: paired sources blend under compression

Iona tries to save time by reading both sources with minimal notes. Later she cannot remember which source used statistical evidence and which used personal testimony. She rereads both. A compact two-line source identity adds less than a minute during first reading and saves several minutes during comparison. The speed gain comes from better separation, not less structure.

109. Practice Lab: graph reading resets the whole problem

Leonie answers a prose question, then treats the graph as an entirely new task. She rereads the prompt and surrounding paragraph from scratch. Training teaches mode carryover: keep the claim active, ask what variable the graph can confirm or challenge, read axes and group, return to the same question state. Switching cost falls.

110. Practice Lab: answer speed improves when thinking ends before writing

Maren often writes, deletes and rewrites because she begins before deciding the answer structure. For one week she must pause briefly to state the logical skeleton: evidence + inference, or A + B on axis. Writing then becomes nearly continuous. The pause adds seconds but removes much larger rewriting cost. Faster output begins with a more stable pre-writing state.

111. Practice Lab: one-mark answers become mini-essays

Iona’s correct responses are consistently long because she adds a second explanation “just in case”. Time review shows that these additions rarely earn marks and sometimes introduce contradictions. She adopts a completion rule: one required point, one necessary bridge, stop. The section becomes faster without changing reading rate at all.

112. Practice Lab: creative paraphrase is slowing the summary

Leonie spends time replacing nearly every source word with a sophisticated synonym. Some replacements distort meaning. She switches to proposition-first paraphrase: preserve actor, relation, condition and degree, then express naturally. Her summary becomes both faster and more accurate because lexical creativity no longer competes with semantic fidelity.

113. Practice Lab: inference remains slow after evidence is found

Maren reaches the correct evidence corridor quickly but continues generating motives. Her problem is not search. Training compares the strongest interpretation with alternatives by assumption cost. Once one answer explains the evidence with fewer unsupported assumptions, she writes and moves. Inference time falls sharply while accuracy stays stable.

114. Practice Lab: tone vocabulary search is the bottleneck

Iona knows the writer is negative but spends ninety seconds choosing among ten adjectives. She begins with dimensions—critical, degree moderate, certainty high, not hostile. The evidence narrows the label to “firmly critical”. Tone selection becomes faster because the search space is structured before vocabulary enters.

115. Practice Lab: purpose becomes faster after scale is identified

Leonie rereads the whole article to answer why one anecdote is included. The question is local. She learns to inspect the sentence before, the anecdote and the sentence after. The function becomes clear: personalise an abstract problem before wider evidence. The next local-purpose question is solved in a much smaller corridor.

116. Practice Lab: comparison speed rises when the axis comes first

Maren previously reread both sources broadly. Now she names the axis “confidence in expansion” before searching. Source A evidence and Source B evidence become easier to locate because irrelevant details disappear. Total comparison time falls even though each selected piece of evidence is read carefully.

117. Practice Lab: evaluation becomes fast after criterion selection

Iona used to ask vaguely whether evidence was “good”. A claim about national prevalence now triggers one criterion: representativeness. She checks sample and scope instead of reviewing every source feature. Evaluation becomes faster because the criterion matches the claim.

118. Practice Lab: summary writing starts later but finishes earlier

Leonie used to draft immediately and rewrite when duplicates appeared. She now spends a short selection phase building distinct in-category propositions. The first sentence is written later, but the entire summary finishes earlier because the point architecture is already stable. Apparent early speed had been creating rework.

119. Practice Lab: reduced annotation preserves search quality

Maren’s six-colour system adds several minutes. A matched fresh-text test removes everything except paragraph jobs, source labels and major turns. Evidence-search accuracy stays unchanged. The old extra marks were not carrying retrieval value. The reduced system is retained as a genuine safe speed gain.

120. Practice Lab: reduced annotation goes too far

Iona copies Maren’s leaner system but removes source labels that she personally relied on. Paired-text errors increase. The compression is reversed partly. Speed architecture must be learner-specific: remove only markings whose retrieval value is low for that reader.

121. Practice Lab: a faster first read lowers map quality

Leonie reduces her reading time by twenty per cent. Map review shows she now misses final stance and one major turn. Later global answers fall. The compression ladder backs up to the previous step. Safe speed ends where structural information begins disappearing.

122. Practice Lab: a faster first read keeps map quality stable

Maren shortens paragraph labels and stops highlighting examples. Her first-read time falls while movement, source identity and final stance remain intact. Reread count also falls slightly. This is a genuine gain: lower time, stable representation, no hidden repair cost.

123. Practice Lab: syntax support fades successfully

Iona initially brackets every clause in dense sentences. Over weeks, the visible marks shrink to one underline under the main proposition, then disappear. Fresh sentences remain accurate. The same parsing operation is still happening, but with less external and conscious cost.

124. Practice Lab: syntax support fades too early

Leonie removes clause marking after one successful session. On fresh technical prose she begins attaching conditions to the wrong claim. The support is restored partially. Fading should follow delayed transfer, not immediate success.

125. Practice Lab: vocabulary triage survives an unfamiliar topic

Maren meets a passage about a field she has never studied. Several nouns are unknown, but the passage defines the key mechanism. She resolves the one term central to the causal chain and leaves the rest as labels. Her map remains accurate. This is transferable lexical triage rather than topic familiarity.

126. Practice Lab: background knowledge creates premature speed

Iona reads a familiar policy topic rapidly because she expects the usual argument. The writer takes an unusual position. She answers from prior knowledge. Training includes familiar topics with deliberately different conclusions, forcing her to verify stance even when orientation is easy.

127. Practice Lab: familiar syntax frees attention for reasoning

After repeated work with concession structures, Leonie no longer stops to untangle “Although X, the writer nevertheless argues Y”. She identifies granted point and surviving claim almost immediately. The saved effort becomes available for the actual question. Language familiarity creates speed without skipping meaning.

128. Practice Lab: discourse patterns become a prediction aid

Maren recognises a problem → failed response → revised solution movement. She predicts that the final section will probably narrow the proposal. The prediction makes later reading easier, but she checks rather than assumes. When the text instead rejects the solution, she updates immediately. Prediction accelerates because it remains provisional.

129. Practice Lab: forced uniform pace damages negation

Iona tries to maintain a constant brisk rate and reads “did not establish” as “established”. The answer is fast and wrong. Her speed plan adds protected micro-slowdowns around negation and conditions. Total time barely changes, but logical accuracy improves sharply.

130. Practice Lab: selective slowdown prevents a cascade

Leonie spends an extra fifteen seconds on a paragraph where the writer concedes one point but preserves the main claim. That small slowdown stabilises purpose, tone and evaluation answers later. The local cost saves several minutes of correction. Not all slowdown is inefficiency.

131. Practice Lab: mixed question types expose cue dependence

Maren is fast on blocked inference drills but slows when inference, literal detail and purpose are mixed. She was relying on the exercise sequence to identify the operation. Training shifts toward interleaving. Command recognition becomes part of the speed skill.

132. Practice Lab: command paraphrase becomes automatic

Iona initially spends time translating unfamiliar question wording. Practice groups several surface forms under the same job—cause, inference, purpose, evaluation. Later she recognises the operation quickly even when the wording changes. Answer speed improves before evidence search begins.

133. Practice Lab: answer skeletons speed writing without making prose mechanical

Leonie uses logical slots rather than memorised sentence starters. For an inference, she knows the answer needs clue + smallest supported meaning. Her actual wording varies naturally. Writing becomes faster because structure is familiar while language remains responsive to the passage.

134. Practice Lab: memorised phrasing slows the learner

Maren forces a favourite sentence frame onto a question it does not fit, then rewrites it. The template has become a detour. Training keeps only the underlying logic and removes compulsory wording. Faster response comes from flexible structure, not recitation.

135. Practice Lab: fast wrong tone answers reveal a rule problem

Iona answers “hostile” instantly whenever criticism appears. More time would not help because the underlying rule is wrong. Contrast cases distinguish cautious, sceptical, critical and hostile. Only after the concept is repaired does speed become desirable again.

136. Practice Lab: slow correct comparison becomes a compression target

Leonie always gets paired-source comparison right but rereads both sources fully. Her source maps and axis choices are already accurate. Training removes the full reread and requires local retrieval from mapped regions. Accuracy stays high while time falls. High-cost correctness yields a safe improvement opportunity.

137. Practice Lab: low-confidence correct answers keep being reopened

Maren’s inference answers are usually correct, but she checks them repeatedly because another interpretation is imaginable. Her stop rule becomes objective: evidence sufficient, scope matched, no contradiction. Confidence may remain moderate. The repeated checking disappears and total time falls.

138. Practice Lab: speed gain disappears after a week

Iona reduces answer time immediately after a lesson on response length. A week later the old long answers return. The gain was cue-dependent. Spaced fresh-text practice is added until concise completion survives delay. Transfer, not same-day speed, becomes the success criterion.

139. Practice Lab: speed survives unfamiliar genre

Leonie learns proposition-first paraphrase in informational texts. A later narrative inference requires the same fidelity to actor, relation and time frame. Her answer remains concise and accurate. The surface has changed, but the underlying compression skill transfers.

140. Practice Lab: short passages hide a navigation weakness

Maren appears fast on 250-word passages. On 900-word texts, evidence search expands because she relied on memory rather than structure. Training increases length while preserving compact mapping. Speed returns as indexing replaces memory dependence.

141. Practice Lab: longer passages do not slow the learner proportionally

Iona’s passage becomes fifty per cent longer, but her total time rises only modestly because much of the additional material is examples whose function is obvious. She does not process every extra word at maximum depth. Variable resolution prevents length from translating directly into equal time growth.

142. Practice Lab: dense short prose is slower than long narrative

Leonie initially thinks something is wrong because a short argument takes longer than a longer story. Analysis shows the short text contains several conditions, source shifts and abstract comparisons. The correct response is not to force narrative speed. Density, not word count, explains the local slowdown.

143. Practice Lab: digital medium exposes a switching-cost bottleneck

Maren reads and reasons normally but loses time moving between passage, graph and answer fields. A medium-specific drill teaches stable screen positions, flag use and minimal scrolling. The intellectual process is unchanged. Speed improves because mechanical switching is cheaper.

144. Practice Lab: handwriting is not the main problem

Iona blames slow handwriting. Timing the process shows most delay occurs before the pen moves because she is still choosing evidence and wording. Once answer skeletons are stabilised, writing becomes continuous enough that total time improves without trying to physically write faster.

145. Practice Lab: broad reading creates speed over several weeks

Leonie reads varied essays, narratives and science journalism outside exam drills. Over time, reporting verbs, concession patterns and abstract nouns require less conscious effort. Timed comprehension becomes faster even though no special “speed technique” was added. Language familiarity has lowered processing cost.

146. Practice Lab: repeated full papers do not fix one syntax bottleneck

Maren completes many mocks but still stalls on embedded conditions. More full papers repeat the problem without isolating it. A short syntax module improves the local operation, after which full-section speed rises. The repair occurs below the level where the symptom was visible.

147. Practice Lab: faster checking creates more harmful changes

Iona rushes the checking phase and changes answers based on feeling rather than named defects. Final time is shorter but score variance rises. 110’s risk hierarchy is restored. Speed compression is allowed only in low-value polish, not in evidence-based revision criteria.

148. Practice Lab: checking compresses safely

Leonie’s ten-minute checking routine includes long sequential rereading. Training reduces it to blanks, partials, flagged high-value items, personal risks and mechanics. Marks recovered remain stable while checking time falls. This is safe compression because the low-yield layer was removed.

149. Practice Lab: the learner stops chasing a lower time once quality is stable

Maren reaches a point where further first-read compression repeatedly harms map quality. Instead of treating the plateau as failure, training shifts to another bottleneck: answer length. Overall section time still improves. Speed engineering chooses the next reducible cost rather than attacking a stable efficient operation.

150. Practice Lab: the mature reader feels slower in the right places

Iona notices she still slows at a dense concession, a source-attribution shift and a key conditional sentence. Yet her total section time is lower than before because easy material, retrieval and answer construction are more efficient. Mature speed is not a feeling of constant rushing. It is selective precision embedded inside a cheaper overall system.

Technical Appendix A: The speed–accuracy diagnostic matrix

The same complaint—“I cannot read fast enough”—can be produced by very different mechanisms. A diagnostic matrix separates them before training begins. The first column is the visible symptom, the second is the likely bottleneck, the third is the control variable that must remain accurate, and the fourth is the smallest useful test. The purpose is not to turn reading into bureaucracy. It is to prevent a general speed prescription from being applied to the wrong problem.

Slow first read + low reread count + high accuracy: the learner may be over-resolving the passage but building a strong model. Test a lighter annotation system or earlier stop rule. Preserve passage movement, voice and final stance as control variables. Fast first read + high reread count: the learner may be under-reading. Increase structural representation rather than demanding more speed. Normal first read + slow questions: inspect evidence retrieval, inference closure and answer construction. Fast answers + low accuracy: inspect premature closure, keyword matching, scope loss or conceptual misconceptions before any compression.

Slow only on unfamiliar topics: test lexical triage and relation-first reading. Slow only on arguments: inspect claim architecture, concessions and source ownership. Slow only on narratives: inspect chronology, viewpoint and distributed motive evidence. Slow only on paired texts: inspect source separation and comparison-axis selection. Slow only late in the paper: the speed problem may be fatigue resistance rather than basic reading rate.

Slow writing after fast reasoning: inspect answer length, paraphrase habits and wording perfectionism. Long pauses before writing: answer contracts may be unstable. Repeated checking after correct answers: confidence calibration may be the bottleneck. Large score drops when time is compressed slightly: the underlying process may not yet be automatic enough for speed training. Stable score but no further time improvement: the learner may have reached the efficient floor of the current operation and should target another bottleneck.

The matrix should always lead to a falsifiable test. If the hypothesis is “over-annotation”, remove selected markings on fresh texts and compare total passage cost. If the hypothesis is “weak syntax processing”, give dense sentences outside a full passage and measure local repair. If the hypothesis is “confidence-based overchecking”, preserve first-pass and final answers and compare change yield. Speed diagnosis improves when every theory has a way to be wrong.

Technical Appendix B: The total-cost model

Useful comprehension speed can be represented as a chain of costs rather than a single stopwatch number:

TOTAL COST = ORIENTATION + MODEL BUILDING + LOCAL REPAIRS + QUESTION RECOGNITION + RETRIEVAL + REASONING + EXPRESSION + CORRECTION CREATED BY EARLIER INSTABILITY.

The final term is crucial. A learner may cut orientation and model-building time but increase repair later. Another may spend slightly more time before writing and eliminate several sentence rewrites. Another may reduce answer wording and create a larger reserve without changing reading speed at all. Total cost exposes these trade-offs.

The model also explains why two students with the same total time need different training. Student A spends heavily on the first read but retrieves quickly. Student B reads quickly but repeatedly scans the passage. Student C is efficient at reading and retrieval but overthinks inference. Student D reasons quickly but writes long answers. Training should compress the expensive term, not the term that happens to be called “reading”.

During practice, the cost model can be estimated roughly rather than timed to the second. Use natural boundaries: first read completed, evidence located, answer begun, answer completed. The purpose is to identify disproportionate cost, not to create laboratory precision. Once the bottleneck is clear, external measurement can fade.

A healthy compression creates one of three patterns. First, one term falls while all others remain stable. Second, a small rise in an early term produces a larger fall downstream—for example, stronger mapping reduces repeated search. Third, automation reduces conscious cost without altering visible procedure—for example, syntax relations are recognised faster even though the reader still processes them accurately. An unhealthy compression shifts cost or creates error.

Technical Appendix C: Safe-compression hierarchy

Not all comprehension activity is equally safe to compress. A practical hierarchy starts with behaviours that often contain high waste and moves toward behaviours where aggressive compression carries larger accuracy risk.

• Usually safest to reduce first: decorative annotation, repeated rereading without a question, stylistic rewriting after the scoring content is complete, long quotations, redundant answer explanations, repeated checking for reassurance, full-text searches after a usable evidence corridor exists.
• Safe to reduce after training: visible syntax marking, written passage-map notes, explicit answer skeletons, long vocabulary-analysis routines, blocked question-family cues. These supports can fade as the underlying decisions become automatic.
• Compress cautiously: passage orientation, evidence verification, inference reasoning, comparison-axis selection, summary point selection. These operations carry essential information and should shrink only after accuracy is stable.
• Protect strongly: negation, conditions, quantifiers, source ownership, major structural turns, chronology boundaries, final stance, graph units and groups, explicit multi-part demands. The time cost is small relative to the damage from missing them.

The hierarchy prevents an intuitive but dangerous strategy: removing the hardest cognition first. Students under pressure often skip passage modelling, evidence checks or conditions because these feel like “slow thinking”, while continuing low-value behaviours such as overlong writing. The order should be reversed. Remove waste before removing meaning-bearing control.

The hierarchy is also learner-specific. A source label may be low-cost and indispensable for Iona but unnecessary for Maren on a simple one-source text. A visible answer skeleton may be redundant for an advanced learner but essential for someone who under-answers. Compress the support only when the supported operation can survive without it.

Technical Appendix D: Reading-resolution matrix

One of the strongest speed controls is choosing the correct reading resolution for the current job. The learner should not use maximal close reading everywhere. Nor should she skim everywhere. Five modes are useful.

D1. Orient

Goal: establish topic, voice, broad purpose and likely text family. Resolution: low to moderate. Risk: premature stance assumption. Stop when the reader has a provisional frame that can be updated.

D2. Map

Goal: identify paragraph jobs, major movement, important turn and final state. Resolution: moderate, with local slowdowns at structural control points. Risk: over-annotation or treating examples as equal to claims. Stop when the passage is navigable.

D3. Close-read

Goal: resolve a dense sentence, device, condition, source relation or difficult proposition. Resolution: high but local. Risk: allowing close reading to spread unnecessarily. Stop when the controlling relation is stable.

D4. Search

Goal: locate the smallest evidence corridor for a known question job. Resolution: selective. Risk: lexical scanning based only on repeated keywords. Stop when sufficient evidence is found, not when every relevant sentence has been collected.

D5. Verify

Goal: test one possible defect in an answer. Resolution: narrowest. Risk: turning verification back into full analysis. Stop after the named defect is cleared or corrected.

Training should practise transitions among modes. A learner may orient quickly, map moderately, close-read one crucial sentence, search locally for evidence and verify one scope word. Speed comes partly from not using the wrong resolution for the job. Re-reading the entire text at close-reading resolution for every question is expensive because the mode never changes.

Technical Appendix E: Syntax-repair protocol

When a sentence resists ordinary reading, use a bounded repair rather than repeated full rereading. Step one: locate the main subject and main verb or, more broadly, the core proposition. Step two: identify relationship markers—although, because, unless, which, while, rather than, despite. Step three: attach each condition or concession to the proposition it modifies. Step four: restate the sentence in simpler internal language. Step five: return immediately to the passage.

The repair should remain local. If the learner starts diagramming every grammatical label, the method has become too heavy. The objective is semantic stability: who or what is doing what, under what condition, despite what qualification, with what result? Grammar terminology is optional if the relationship can be recovered accurately.

Compression training fades the protocol. Initially, the learner may mark clauses. Later, she identifies the core proposition mentally. Finally, common structures become recognised as chunks. The speed gain is demonstrated when dense sentences no longer trigger repeated rereads and later answers remain accurate.

Some sentences should still stay slow. If a question explicitly targets wording or the sentence carries the central qualification of the passage, a careful reread is rational. The syntax protocol makes difficult sentences cheaper; it does not require all sentences to become equally fast.

Technical Appendix F: Vocabulary-triage protocol

When an unfamiliar or unstable word appears, classify it before investing time. Category 1: nonessential label. The sentence relation remains clear without exact meaning; continue. Category 2: locally inferable. Grammar, contrast, example or definition gives enough evidence; infer briefly. Category 3: structurally important. The word controls the main claim, condition or question; resolve more carefully. Category 4: examined target. Use the full vocabulary-in-context routine appropriate to the question.

This protocol protects speed from the assumption that every unknown word deserves equal attention. It also protects accuracy from the opposite assumption that unknown words can always be ignored. The controlling question is functional: what role does the word play in the current passage and current task?

Broad reading outside examinations gradually changes the distribution. Words that once required Category 2 or 3 processing become automatic. That is one reason long-term vocabulary and reading development create examination speed without any new stopwatch technique.

Technical Appendix G: Passage-map compression ladder

Stage 1 — full external map: short phrase beside each paragraph plus major turn and source identity. Stage 2 — compressed map: only paragraph job and turn. Stage 3 — selective external map: visible marks only for turn, source and unusual dependency; routine paragraph jobs remain mental. Stage 4 — mental map with anchor marks: the learner retains structure internally and marks only difficult locations. Stage 5 — adaptive mapping: dense or multi-source texts receive more external support; straightforward narrative may receive almost none.

Advancement requires evidence. The learner should still retrieve the correct evidence regions, state whole-text movement and answer global questions after delay. If performance falls when a stage is faded, return to the previous support level. Compression should remove the visible map only after the internal map can carry the work.

The final stage is not “no annotation”. It is annotation chosen by need. Mature readers may write more on a complicated source set and almost nothing on a clear narrative. Speed comes from adapting the support to the text rather than following one ritual.

Technical Appendix H: Answer-construction compression ladder

Stage 1: write a complete answer skeleton before prose. Stage 2: use two or three keywords representing logical slots. Stage 3: state the skeleton orally or mentally. Stage 4: recognise familiar answer contracts automatically and plan explicitly only for unusual tasks. Stage 5: under mixed examination conditions, generate complete concise responses without visible planning while still preserving scope and component count.

At every stage, completeness controls progression. If the learner begins omitting one side of comparisons or one part of explanation questions, visible planning has faded too soon. A faster answer is not a successful answer if the skeleton disappeared before writing.

Response-length control can be trained at the same time. After marking, highlight the words that actually contribute scoring content. Repeatedly seeing long neutral stretches helps the learner develop an internal sense of when an answer has finished. That sense is one of the largest speed gains available to strong but over-elaborate students.

Technical Appendix I: Question-family speed-risk matrix

• Direct detail: usually compressible through local retrieval and concise expression. Main risk: unnecessary inference or broad search.
• Inference: compress evidence search and alternative generation, but protect sufficiency and certainty calibration. Main risk: premature closure or possibility inflation.
• Vocabulary in context: compress dictionary-style deliberation; protect syntax and local relation. Main risk: familiar-meaning bias.
• Language effect: compress technique cataloguing; protect wording → local meaning → effect relation. Main risk: generic effect labels.
• Tone: compress adjective search through dimensions; protect degree and local/global scale. Main risk: extreme labels.
• Purpose: compress whole-text rereading when task is local; protect scale and structural role. Main risk: generic purpose formulas.
• Structure: compress feature inventory; protect before → move → after. Main risk: naming devices without movement.
• Comparison: compress broad source rereads through shared axis and source maps; protect symmetry and source ownership. Main risk: mismatched dimensions.
• Evaluation: compress generic “strength” discussion by choosing the controlling criterion; protect claim–criterion fit. Main risk: formulaic reliability claims.
• Summary: compress rewriting through staged selection and faithful paraphrase; protect distinctness, category and conditions. Main risk: writing before point architecture is stable.

The matrix shows that “faster answering” means different things across question families. A single universal technique is unlikely to help. Each family has a specific waste source and a specific accuracy variable that cannot be sacrificed.

Technical Appendix J: Passage-family speed architecture

Narrative: accelerate through straightforward chronology and low-stakes description; slow at time shifts, viewpoint changes, recurring motifs and decision points. Preserve who knows what and when. Argument: accelerate once paragraph roles are recognised; slow at concessions, counterarguments, conditions and narrowed conclusions. Preserve claim architecture. Informational: move lightly through repeated examples; slow at definitions, mechanisms, cause chains and qualifications. Preserve hierarchy and relation.

Paired sources: invest early in compact source identity; accelerate source-specific retrieval afterward; slow at comparison-axis formation and any apparent contradiction. Preserve ownership. Multimodal: reduce switching cost by carrying the active claim across modes; slow long enough to establish axes, units, group and evidential job. Preserve what each mode can legitimately prove.

The architecture is flexible because texts can combine families. An informational article may contain narrative anecdote and argumentative conclusion. The speed rule follows the local function, not the document label. Mature readers adjust resolution as the discourse job changes.

Technical Appendix K: Paired-source and multimodal switching protocol

Switching between sources can consume time even when each source is understood well. A compact protocol prevents a full cognitive reset. First, preserve the active question job. Second, preserve the comparison or evidential axis. Third, identify what the new source can contribute. Fourth, read only enough of the new mode or source to fulfil that contribution. Fifth, return to the answer state without reconstructing the entire problem.

For paired prose, this may look like: “Question asks why A and B differ in confidence → A confidence already mapped → open B only for evidence standard → compare.” For prose plus chart: “Claim concerns magnitude → prose gives cause → chart needed only for scale → read variable, unit and group.” The state survives the switch because the learner knows why she is moving.

Practice should include unnecessary-switch traps. Give a question whose answer can be completed from Source A alone even though Source B is visually nearby. The learner should resist switching simply because another source exists. Then give a question that genuinely requires both. Speed improves when source transitions are demand-driven rather than habitual.

Digital interfaces magnify switching cost. Scrolling, tabs or split panels can erase spatial memory. Train stable landmarks and the platform’s navigation. The intellectual protocol remains the same, but the mechanical route should become cheap enough that it does not consume the attention required for source comparison.

Technical Appendix L: Teacher protocol for speed–accuracy training

1. Establish the accuracy floor. Confirm the learner can perform the target operation correctly without aggressive timing.
2. Measure total cost. Record enough boundaries to identify where time is actually being spent.
3. Select one bottleneck. Avoid changing multiple strategies at once.
4. Name the protected information. For mapping: turn, voice, final stance. For inference: evidence sufficiency and scope. For comparison: source ownership and shared axis.
5. Remove obvious waste first. Repeated rereading, redundant annotation, answer inflation or reassurance checking may offer immediate gains.
6. Build a compression ladder. Reduce support or time gradually rather than demanding a dramatic jump.
7. Test on fresh material. Same-text speed can be memory.
8. Track the first accuracy defect. Stop compression when meaning-bearing information begins to fail.
9. Retest after delay. Same-day fluency is not enough.
10. Recombine under mixed conditions. The faster operation must survive other question types, fatigue and realistic examination state.

The teacher should avoid praise such as “Good, you finished faster” until the accuracy controls are checked. Better feedback is: “You cut ninety seconds and still preserved the concession, source identity and answer completeness.” That sentence tells the learner what made the speed legitimate. Similarly, if speed rises but scope fails, say which control disappeared first. The learner should understand that accuracy is not a general feeling; it is a set of preserved information.

Do not use the stopwatch as punishment. Timing is a measurement instrument. When a learner crosses the current speed boundary, the correct response is diagnosis, not pressure. The boundary tells the teacher what has not yet automated. Restore the smallest support, repair the mechanism, then test compression again.

Teacher judgement also matters when the learner is already efficient. Some students need permission to stop compressing. If a dense comparative question reasonably needs time and further reduction produces repeated errors, move to another bottleneck. The objective is a lower total section cost, not a record time for every individual operation.

Technical Appendix M: Self-study protocol

A self-studying learner can run a simple speed experiment on two or three fresh passages. First record the current routine: first-read time, approximate reread count, longest question, and any answer changes caused by initial misunderstanding. Mark the work. Then choose one bottleneck only. Examples: shorten annotation, use one inference stop rule, reduce one-mark answer length, or triage unfamiliar vocabulary.

On the next fresh passage, change only that behaviour. Record the same measures. Ask: Did total time fall? Did accuracy stay stable? Did rereading rise? Did a specific accuracy defect appear? Did the learner merely shift the cost elsewhere? If the result is mixed, keep the experiment small instead of turning the new method into a permanent rule.

A useful self-study log has six fields: bottleneck, compression attempted, protected variable, fresh-text result, delayed result, decision. The final decision is keep, modify, or reverse. This prevents strategy accumulation. A learner should not end up carrying every technique ever tried.

Self-study learners should be especially cautious with raw words-per-minute goals. If a target encourages skipping structural turns or conditions, it is misaligned with comprehension performance. Use passage-completion cost and answer accuracy instead. The best speed metric is the one closest to the actual scoring task.

Technical Appendix N: Parent support protocol

Parents can help by replacing “You read too slowly” with operational questions. “Where do you lose most time?” “Do you reread the whole passage?” “Are your answers too long?” “Does unfamiliar vocabulary stop you?” “Do you still get the same marks when you go faster?” These questions direct attention toward mechanisms instead of identity.

Do not compare raw reading speed between children. One may be reading denser material, another may be taking shortcuts, and another may be strong enough that speed is not the active bottleneck. Compare the learner with her own earlier total cost and accuracy on varied fresh texts.

Parents can also support the broad reading that creates long-term fluency. Varied prose builds vocabulary, syntactic familiarity and discourse expectations without the stress of constant timing. Examination speed is partly the visible result of years of language exposure. Not every improvement needs to look like a timed worksheet.

If a child becomes anxious about the clock, avoid increasing time pressure indiscriminately. Ask whether the current task is training accuracy, compression or full performance. Some sessions should be comfortably timed so the mechanism can stabilise. Pressure belongs later in the progression.

Technical Appendix O: One-week speed repair cycle

Day 1 — Baseline. Use one or two fresh passages and identify the largest unnecessary cost. Day 2 — Isolate. Work only the target operation without severe timing. Day 3 — Compress slightly. Reduce support or time while preserving the named accuracy variable. Day 4 — Vary surface. Use a new topic or genre with the same mechanism. Day 5 — Interleave. Place the target among other question types. Day 6 — Delay and retest. Use fresh material without reminders. Day 7 — Review total cost. Decide whether to keep the gain, modify it or restore the previous boundary.

One week should not chase several independent speed goals. If annotation, inference and writing are all changed, a better result cannot be attributed. Keep the experiment narrow. Once one gain transfers, it joins the ordinary system and the next bottleneck can be selected.

If the learner becomes less accurate during Day 3, the week is still useful. The compression test has located the boundary. Return to Day 2 conditions and strengthen automaticity before trying again. Failure under compression is information about the current system, not proof that the learner cannot become faster.

Technical Appendix P: Six-week speed build

Week 1 — Map the cost. Establish current first-read, reread, question, writing and checking patterns. Choose one or two high-leverage bottlenecks. Week 2 — Remove obvious waste. Reduce annotation excess, repeated broad rereading and answer inflation where present. Do not yet compress meaning-bearing operations aggressively.

Week 3 — Build component fluency. Work syntax repair, vocabulary triage, command recognition or another diagnosed component. Week 4 — Compression ladder. Fade visible support or reduce time in small steps, using fresh matched texts and paired accuracy metrics. Week 5 — Transfer. Change topic, genre, passage length, question wording and modality. Week 6 — Recombine. Put the faster operation into realistic timed comprehension sections and verify that the gain creates usable reserve rather than downstream repair.

The six-week outcome should be specific. “Reads faster” is vague. Better outcomes are: first-read time reduced while map quality remains stable; full rereads fall from three to one; one-mark answer time falls because response length is controlled; inference closure occurs after sufficient evidence instead of three alternative readings; paired-source comparison no longer requires full source rereads.

If no safe speed gain appears, inspect the language base. Weak vocabulary, syntax or reading experience can create a genuine processing floor that strategy cannot bypass. Shift more training into language development and broad reading rather than escalating stopwatch pressure.

Technical Appendix Q: Twelve-week calibration and maintenance arc

Weeks 1–2: baseline and total-cost diagnosis. Weeks 3–4: remove low-value repetition and excessive external support. Weeks 5–6: automate one or two language or reasoning bottlenecks. Weeks 7–8: compress carefully and establish the current safe boundary. Weeks 9–10: transfer across text families, density and modality. Weeks 11–12: integrate with realistic section pacing, checking reserve and fatigue.

Maintenance should be lighter. Occasionally audit total cost on a fresh passage. If broad rereading returns, refresh the map system. If answers lengthen again, revisit response contracts. If scope begins failing under speed, back off the compressed pace. The system should be maintained by symptom, not by repeating the entire programme permanently.

A mature learner may continue becoming faster through ordinary reading development even after formal speed training ends. Vocabulary grows, common syntactic structures become automatic and discourse patterns become familiar. The twelve-week programme should therefore establish good compression habits, not pretend to set a permanent maximum reading rate.

Technical Appendix R: Anti-false-speed audit

• Did first-read time fall while full rereads increased?
• Did the learner finish earlier by leaving subparts incomplete?
• Did quantifiers, conditions or certainty become less precise?
• Did source-attribution errors increase?
• Did the learner answer more from keywords or topic familiarity?
• Did high-confidence wrong answers become more common?
• Did answer changes increase because the first pass was unstable?
• Did map quality fall even though the learner felt faster?
• Did the learner stop resolving unfamiliar words that actually controlled the claim?
• Did the learner force one pace through both low-density and high-density text?
• Did the final checking reserve grow only because important verification disappeared?
• Did score variance rise across comparable mocks?

Several “yes” answers indicate borrowed speed. Restore the first operation that failed, not necessarily the entire old routine. If only source identity weakened, strengthen source mapping while keeping shorter answer length. If only conditions disappeared, restore micro-slowdowns around logical control words. Precision repair preserves the gains that were genuine.

False speed is especially dangerous because it can feel successful. The learner sees an earlier finish and may attribute later errors to carelessness. The audit forces a different question: what information stopped surviving the faster process? Once that is known, the speed boundary becomes trainable.

Technical Appendix S: Full worked walkthrough — compressing an argumentative passage safely

Consider a long argument about whether a city should expand a public transport scheme. The learner’s baseline is accurate but slow. The first read takes nine minutes. She writes notes beside every paragraph, underlines nearly every statistic and rereads the introduction before each global question. A full question set takes thirty-five minutes. Her score is strong. The task is not to make her skim. It is to discover where the thirty-five minutes contain unnecessary cost.

Baseline decomposition. The first read shows good comprehension. The map contains twelve long notes for twelve paragraphs. Later review shows she uses only five of them: opening claim, pilot evidence, cost concession, equity counterargument and narrowed conclusion. She rereads globally three times, largely because the dense notes are difficult to scan. Her one-mark answers are also long. The high-cost operations are annotation density, retrieval and answer length; the important claim architecture itself is not weak.

Compression 1 — map. On a fresh argument, Maren writes only a two- or three-word job for each major section and circles the primary turn. Examples receive no separate notes. First-read time falls modestly. Crucially, after reading she can still state the writer’s opening position, concession and final narrowed recommendation. Global-question accuracy remains intact. The change is kept.

Compression 2 — retrieval. For each question, she predicts the evidence region from the map before touching the text. She is not allowed to restart from paragraph one unless local widening fails. Full rereads drop from three to one on the next fresh text. Evidence accuracy remains stable. The learner has not become a faster eye mover; she has made the passage cheaper to navigate.

Compression 3 — answer length. Marked scripts show that one-mark direct answers often contain two extra sentences. Maren begins writing to logical slots. One direct detail: proposition, stop. One inference: clue-linked meaning, calibrated certainty, stop. Comparison: A + B on shared axis. Answer time falls without a completeness loss.

Protected slowdowns. The programme does not speed up everything. Maren continues to slow at the paragraph beginning “Although the pilot reduced waiting time…” because it contains a concession and condition. She also reads the final recommendation carefully. Those local slowdowns protect several later questions. Their seconds are not treated as waste.

Transfer test. A week later she receives an unfamiliar argument about conservation policy. Surface vocabulary changes. The text contains a similar claim → evidence → concession → revised recommendation movement. Without teacher cues she uses the lean map, local retrieval and concise answer contracts. Total passage cost falls meaningfully while score and scope fidelity remain stable. The speed gain is now more credible because it survived fresh topic, delay and unseen wording.

Failure boundary. When first-read time is compressed further, she begins missing final qualifications. The programme stops reducing the first read. Instead, later work targets faster contextual vocabulary decisions. Overall section time can still improve without crossing the structural-reading boundary. This is the core logic of 120: preserve efficient floors and move to the next reducible cost.

Technical Appendix T: Full worked walkthrough — dense informational and narrative text

Iona has the opposite profile. She reads ordinary narratives quickly but slows dramatically on dense informational prose and unfamiliar domains. Her first response is to practise “speed reading”. The result is poor: conditions disappear, technical words are skipped even when important, and she answers causal questions from surface sequence. The correct intervention begins below whole-passage pace.

Diagnosis. Timed sentence work shows the largest delays occur around embedded conditions, reference chains and unfamiliar technical nouns. When these are supplied in simpler language, her reasoning is fast. The bottleneck is local parsing and lexical triage, not general comprehension ability.

Component repair. Iona practises dense sentences using core proposition → condition → qualifier. She also classifies unfamiliar words by importance. Terms that function as labels can remain unresolved; terms controlling the mechanism receive local context analysis. After several sessions, local repair becomes shorter and she no longer restarts whole paragraphs.

Recombination. A fresh science article contains eight technical terms and several causal claims. Iona resolves two key terms, treats others as placeholders and maps the mechanism: process A changes variable B only under condition C, producing outcome D. She reads the article more quickly not because she understands every noun, but because the relational skeleton is strong enough to carry the questions.

Narrative transfer. Later, a narrative contains a long sentence with a flashback and qualifying clause. The same syntax-repair skill applies. Iona identifies current action, earlier event and the condition linking them. The operation learned on informational prose transfers across genre because the underlying relation is grammatical and temporal rather than domain-specific.

Protected detail. In a later literary passage she encounters repeated imagery. Because recurrence may matter to symbolism, the detail is not treated like an expendable scientific example. Variable resolution follows local function. The reader moves lightly through background description but slows when an image repeats at a structural turn. Speed remains interpretively intelligent.

Outcome. Iona’s average first-read time falls modestly, but the larger gain comes from fewer repair loops. Broad rereads disappear, technical vocabulary no longer freezes the passage and dense sentences are repaired locally. Total completion falls much more than raw reading time. This demonstrates why 120 uses total useful processing cost rather than a headline words-per-minute number.

Technical Appendix U: Handoff to the score-plateau problem

A learner can become faster and still stop improving in marks. The next reserved spoke, SETC-SEC-COMP-130 — Why Reading Comprehension Scores Plateau | When More Practice Papers Stop Working — owns that broader problem. It asks why performance ceases to rise despite continued practice: unresolved first weak links, language-base ceilings, repeated practice without transfer, stale feedback, difficulty mismatches, execution variance, overfitting to familiar paper types and limits that are not fundamentally about speed.

The boundary is clean. 120 owns the engineering of lower processing cost with protected accuracy. If speed training succeeds but the learner still loses the same inference or evaluation marks, the plateau belongs elsewhere. If scores rise until the learner reaches time pressure and then stop, 120 may still be part of the diagnosis. 130 integrates the broader plateau evidence and decides why more practice volume is no longer producing progress.

151. How SETC-SEC-COMP-120 connects to the lane

This page owns comprehension-specific speed–accuracy engineering: locating processing bottlenecks, reducing low-value cost, varying reading resolution, compressing passage mapping and answer construction, protecting meaning-bearing slowdowns, and proving speed gains on delayed fresh texts. It receives the passage-model foundation from 020, clock context from 030, evidence-search architecture from 050, transfer methodology from 090, mock evidence from 100, and final-reserve controls from 110.

Return to SETC-SEC-COMP-000 for the complete lane. The next reserved spoke is SETC-SEC-COMP-130, the comprehension score-plateau problem.

152. Final principle: faster means less wasted work, not less understanding

The strongest speed gains do not come from asking the eyes to race across every sentence. They come from building a cheaper comprehension system. The learner stops repeating completed work, recognises familiar relations as larger units, uses a passage map as an index, changes reading resolution according to risk, closes reasoning when evidence is sufficient, writes only what the scoring job requires and preserves small words that control large meanings.

Maren protects structure and complete expression. Iona protects evidence, source and scope. Leonie protects the economics of the process: where seconds produce value, where they are being wasted, and where a small slowdown prevents a much larger repair. Mature speed is therefore selective. It accelerates what can safely automate and refuses to rush what still carries the meaning of the text.

THE GOAL IS NOT TO READ EVERY PART OF THE PAPER FASTER. THE GOAL IS TO COMPLETE THE RIGHT COMPREHENSION WORK WITH LESS TOTAL COST AND NO LOSS OF THE INFORMATION THAT EARNS THE MARKS.