IELTS Reading’s Paraphrase Engine: Vocabulary That Unlocks Answers

Problem: the passage says one thing, the question says the same thing differently, and exact-word matching leads you to the wrong paragraph. Answer in brief: treat every question as a paraphrase puzzle — learn the five paraphrase moves, infer unknown words from context slots, and choose titles by whole-passage coverage rather than vivid details. This guide builds that engine with drills, models and 12 practice tasks with answers.

Learning outcomes

By the end you will be able to: recognise all five paraphrase moves (synonym, structure shift, category swap, polarity flip with negatives, number/date reformatting) between passage and question; infer unknown words from context slots with checkable confidence; choose passage titles by coverage (rejecting too-narrow and too-broad options); solve list-selection and multi-answer multiple-choice sets with per-letter evidence; and build topic vocabulary sets from practice passages. All guidance follows the source book, IELTS ACADEMIC R 02: The Complete IELTS Reading Workbook (Academic) (Jobin Thomas), whose vocabulary-and-paraphrasing chapters, per-type drill bank and full tests with explanatory keys were verified by full-chapter scan.

The five paraphrase moves with examples

Passages and questions state the same idea in different words — deliberately. Exact-match scanning finds nothing; move-matching finds everything. The five moves:

• 1. Synonym swap: “banned private vehicles” → “curbed emissions”; “collapsed” → “went bankrupt”; “economic downturn” → “recession”. Content words change, function words stay. Build synonym pairs per topic (ban/curb/prohibit; rise/increase/climb).
• 2. Structure shift: active↔passive (“researchers tracked fish” → “fish were tracked”), clause→phrase (“because parents could not read” → “due to parental illiteracy”), word-form change (“pollution showed no effect” → “pollution was ineffectual”). Same proposition, rebuilt grammar.
• 3. Category swap: example↔general (“Labradors and poodles” → “dog breeds”; “2019, 2020, 2021” → “recent years”). Questions generalise passage specifics and vice versa — track up and down the ladder.
• 4. Polarity flip: meaning preserved through negation (“few survived” → “most perished”; “rarely enforced” → “often ignored”). Double-check flips: one missed “not” reverses the verdict.
• 5. Number/date reformat: “two-thirds” → “67%”; “a decade” → “ten years”; “fortnightly” → “every two weeks”. Convert before comparing; never match “most” to “60%” without the passage’s own framing.

Drill: take any solved T/F/NG set and label each answer’s move — answers stop feeling arbitrary once every True carries a named move and every False names its breaking word.

Worked move chains (each: passage fragment → question form → move): (a) “The scheme paid for 23 confirmed kills” → “How many kills were compensated?” — synonym (paid→compensated) plus question-formation; answer “23”. (b) “Wolves thinned elk herds, which had overbrowsed willow” → “Elk overbrowsing damaged willow before wolves returned” — structure shift (relative clause→main clause) with a true temporal inference. (c) “Labradors and poodles aided patients” → “Dog breeds aided patients” — category swap upward. (d) “The cap is a trial measure” → “The cap is not permanent” — polarity-adjacent restatement (trial = not permanent); verify, don’t assume. (e) “Fell by a fifth” → “dropped 20%” — number reformat. Five chains, five verifications: in each case the check is the same — same actors, same time, same scope, only wording changed.

Vocabulary in context: inferring without a dictionary

Unknown words are tested through inference, not recall. The book’s vocabulary exercise models the method: “The new city park project, though ambitious, was met with enthusiasm…” — ambitious most likely means requiring a lot of effort (not merely expensive, not fast), because enthusiasm for green space plus the concessive “though” frames it as scale-effort. Infer in four slots: (1) conjunction frame (though/but/and signals contrast or continuation); (2) grammar slot (adjective before a project noun = scale or cost, narrowed by neighbours); (3) nearby definition (dashes, brackets, “such as”); (4) paragraph direction (problem paragraphs push negative senses).

Confidence discipline: mark each inference A (slot + neighbour agree), B (one clue only) or C (direction only) — spend C-words’ verification time only when they sit inside evidence sentences. After marking, convert C-words into topic sets (the book’s themed passages work this way): word, slot, clue, and one paraphrase pair each. Ten sets like this outperform a hundred flashcards because every entry carries a retrieval context.

Second worked inference (newly written): “The lagoon’s output proved erratic, swinging from surplus to shortfall within weeks.” Erratic most likely means — A. disappointingly low / B. irregular and unpredictable / C. gradually declining. Answer: B. The slot (adjective before a performance noun, glossed by the participial phrase) plus the swing image eliminate steady-decline (C) and pure-low (A, contradicted by “surplus”). Two clues agreeing (slot + gloss) makes this an A-confidence inference — answer and move on. Contrast a C-confidence case: an unknown adjective with no gloss far from any evidence sentence — there, direction-only guessing is allowed solely to avoid blanks, and the word goes straight into the topic set.

The sixth move: who says what (attribution)

Many hard questions turn on attribution, not vocabulary: owners blame delivery restrictions (not the ban), economists add recession (not owners), officials stress trial status (not permanence). The paraphrase may be perfect while the owner is wrong — and owner-wrong fails exactly like fact-wrong. Track three attribution patterns: claim downgrades (passage “owners blamed X” → question “X caused it” stated as fact — upgrade fails); speaker swaps (Nair’s proposal attributed to the ministry — swap fails); consensus inflation (“one study found” → “scientists agree” — single-to-consensus fails). In features sets this is the entire game; in T/F/NG it hides inside otherwise clean paraphrases. The fix is mechanical: underline the owner in the question and the reporting verb in the passage before comparing anything else.

Worked attribution trio (newly written): passage: “Owners blamed delivery restrictions; economists noted a national recession.” (i) “Owners attributed collapse to delivery restrictions.” → True (clean synonym chain, owner intact). (ii) “Delivery restrictions caused the collapse.” → False-or-Not-Given depending on strictness: the passage reports blame, never fact — most keys mark this False as an unattributed upgrade, and where keys say Not Given the reasoning is identical (causation unestablished). Either way it is not True. (iii) “Economists blamed the ban.” → False (speaker swap: economists cited recession, not the ban). Same vocabulary field, three different verdicts — owners decide.

Choosing titles: coverage beats vividness

Choosing-a-title asks for the whole passage’s (or section’s) best heading from a list. The correct title covers the main idea; distractors are too narrow (one vivid detail) or too broad/off-topic. Decide with two questions: “What is this mostly about?” (topic + stance in one line) and “Which title survives every paragraph?” — test each option against each paragraph, discarding any title a paragraph contradicts or ignores. The book’s caffeine passage demonstrates the shape: a title about “morning alertness” is too narrow (one benefit paragraph), “stimulants in modern life” too broad (passage is caffeine-specific), while a title joining mechanism plus balanced effects survives all paragraphs.

Titles differ from headings: headings map each paragraph separately (many answers), titles judge the whole (one answer). A title that fits two paragraphs perfectly but ignores the conclusion is a detail distractor wearing scope clothes — the conclusion test (“does the final paragraph fit this title?”) kills most of them.

List selection and multi-answer multiple choice

List selection (“choose TWO/THREE letters from A–G”) and multi-answer MC need per-letter evidence: each correct letter must independently satisfy the stem, and finding one never validates another. Method: convert the stem into slot requirements (person? reason? figure?), scan each option against the passage separately, and classify every eliminated letter — contradicts / true-but-unasked / unmentioned / partially-true. Partial truths are the designed trap: an option half-supported by the passage fails like any False. For “which TWO factors” sets, expect the two answers in different paragraphs (test-makers split them); two letters from one sentence deserve re-verification, not confidence.

Paraphrase-proofing checklist: a worked application

Run every answer through five lines before banking it: (1) moves named — which of the six carries this answer? (2) breakers hunted — absolutes, dates, owners, numbers checked? (3) owner tagged — whose claim is this, and does the question name them? (4) slot fitted — for completions, does the form match the gap? (5) limit counted — words and units within bounds? Worked on Model 1 Q2: moves — synonym chain (numerous/many, bankrupt/collapsed, recession/downturn); breakers — none (no absolutes, dates match “last year”); owner — unattributed general statement matching an unattributed passage generalisation, clean; slots/limits — n/a (Yes/No). Five lines, thirty seconds with practice, and the arbitrary-feeling verdicts become auditable decisions. Laminate the checklist mentally: moves, breakers, owner, slot, limit.

Worked models: paraphrase chains and title decisions

Labelling: all passages and items below are newly written for this article, adapting the book’s exercise patterns (context-inference MCQ, paraphrase Yes/No, title-choice with narrow/broad distractors). The book’s passages, tests and keys are not reproduced. Teaching material, not official exam content.

Model 1 — Paraphrase chains with move labels (newly written)

Passage (newly written, 141 words): A. The city banned private vehicles in the centre in 2022 to reduce pollution. Annual emissions fell by a fifth within two years. B. Many small businesses collapsed during the same period, which owners blamed on delivery restrictions rather than the ban itself. Economists note the downturn coincided with a national recession, complicating any causal claim.

Questions: 1. “What measure did the city take to curb downtown emissions?” (short answer, ≤3 words.) 2. “Numerous small companies went bankrupt because of last year’s recession.” Same idea as B? (Yes/No.) 3. T/F: “Emissions dropped 20% after the vehicle ban.”

Answers: 1. banned (private) vehicles — synonym chain: curb→reduce, downtown→centre, measure→ban; “banned vehicles” fits, “banned private vehicles” (3 words incl.?) — count: banned(1) private(2) vehicles(3) fits ≤3. 2. No — B attributes collapse claims to owners (delivery restrictions) and adds recession as complication; the statement upgrades contested attribution to fact (owner-claim → fact, the attribution shift). 3. True — number reformat: a fifth = 20%; “within two years” supports “after”. Move labels: Q1 synonym+category, Q2 attribution (a sixth move worth naming: who-says-what changes), Q3 number reformat.

Model 2 — Title decision (newly written)

Passage (newly written, 128 words): Caffeine, the world’s most used psychoactive substance, blocks adenosine and raises alertness; moderate doses cut fatigue while adrenaline lifts heart rate. Students prize pre-exam focus; drivers value night-time vigilance. Yet heavy use brings anxiety, fragmented sleep and dependence, with withdrawal headaches punishing abrupt quitters. Regulators now weigh dose labelling against outright age limits.

Titles: A. “Why students love coffee” / B. “Caffeine: mechanism, benefits and costs” / C. “The dangers of modern stimulants” / D. “How adrenaline works”.

Answer: B. A is too narrow (one user group); C is too broad (stimulants generally) and one-sided (ignores benefits); D is a detail (one mechanism step). Only B survives every paragraph — mechanism, benefits, costs, regulation. Method shown: paragraph-survival test kills A at paragraph 3, C at paragraph 1, D at paragraph 2.

Model 3 — Title decision on an argument passage (newly written)

Passage (newly written, 122 words): Rewilding divides ecologists and ranchers. Yellowstone’s wolf return thinned elk, revived willow and brought back beavers — a trophic cascade its advocates call restoration. Ranchers counter with livestock losses, though both sides accept the 23-kill compensation figure. Researchers caution that Yellowstone’s scale limits generalisation: smaller reserves show weaker cascades, and fenced landscapes block the movements cascades require. The debate, in short, is whether one spectacular success travels.

Titles: A. “Wolves saved Yellowstone” / B. “Rewilding: one success, open questions” / C. “Ranchers versus ecologists” / D. “How trophic cascades work”.

Answer: B. A is one-sided advocacy (ignores dispute + generalisation warning); C captures only the dispute paragraph; D captures only the mechanism. B alone holds success plus limits — topic (rewilding) with stance (qualified). Note the stance test: titles carry attitudes, not just topics — advocacy, dispute-only and mechanism-only titles each fail a different paragraph, while the qualified title survives all four.

Model 4 — Multi-answer multiple choice with per-letter evidence (newly written)

Stem: Which TWO factors complicated causal claims about business collapse? A. delivery restrictions B. national recession C. the vehicle ban D. falling tourism. Passage zone: Model 1’s paragraph B.

Answers: A and B. A: owners blamed delivery restrictions (competing cause). B: economists’ recession coincidence (competing cause). C fails — the ban is the policy, not a complicating factor, and owners explicitly deflect from it. D is unmentioned (tourism never appears). Notes: each letter independently evidenced; C is the true-but-misroled distractor (real element, wrong role); D is the unmentioned-content trap — plausible in life, absent in text, never correct.

Practice bank: 12 tasks with answers

Attempt before opening each answer. All items are newly written.

Common errors and corrections

• Exact-match scanning: hunting question words verbatim. Fix: convert each keyword to 2–3 paraphrases before scanning.
• Unlabelled Trues: “feels right” without a move. Fix: name the move(s) or downgrade to guess.
• Attribution blindness: owner-claims upgraded to facts. Fix: tag who-says-what before grading.
• Vivid-title seduction: choosing the most interesting detail. Fix: paragraph-survival test for every option.
• Letter piggybacking: validating B from A’s evidence. Fix: independent evidence per letter.
• Bare word lists: synonyms without contexts. Fix: every pair stored with its proposition.

Independent task and self-check

Task (new): Take one book passage with its key and annotate every answer with its paraphrase move(s) plus any breaker checked (absolute, date, owner, number). Then build a 10-pair topic set from that passage’s chains and re-solve one question set using only your set as the bridge. Guidance: use the printing-press practice passage first (concrete vocabulary), then a full test passage; time the annotation (aim 15 min) separately from solving. Checklist: every True/Yes carries named moves; every False/No names its breaker; title survived all paragraphs; each list letter independently evidenced; topic set has propositions, not bare words.

Worked annotation sample (format to copy): Q: “Emissions dropped 20% after the ban.” → verdict True | moves: synonym (fell/dropped) + number reformat (fifth/20%) | breakers: date OK (2022+2y supports “after”), scope OK (centre emissions throughout) | owner n/a | slot/limit n/a. One line per answer in a notebook — after twenty lines the moves become reflexes, and unseen passages start resolving the same way.

Study sequence with the book

Week 1: vocabulary + paraphrasing chapters with Exercise 3. Annotate moves on ten solved items daily; start one topic set per domain (environment first, since its chains recur across tests). Week 2: 14-type drill bank, focusing on titles, list selection and multi-answer MC with per-letter evidence; run the proofing checklist on every answer until the five lines fit in thirty seconds. Week 3: printing-press practice test plus time-management and test-day chapters under 20-minute passage timing; practise title decisions with the paragraph-survival test written out. Week 4: full-length Tests 1–2 under 60 minutes with explanatory keys as arbiter; convert every miss into a breaker rule or a topic-set pair. Re-solve one full set aloud naming moves throughout — fluency in move-naming predicts transfer to unseen passages better than any amount of silent re-reading.

FAQs

How many synonyms must I learn? Topics, not totals: 30–50 pairs per frequent domain (environment, health, education, technology) with propositions attached beats 500 bare words.

What if I know none of the words? Work slots and frames — grammar plus direction often answer without any content word. Mark and bank the words after.

Titles or headings first? Titles judge wholes; practise them on complete passages only. Headings map parts; practise them paragraph by paragraph. Never substitute one’s method for the other’s.

Do numbers paraphrase often? Constantly: fractions↔percentages, periods↔dates, rates↔totals. Convert before comparing, every time.

Can two list letters share a paragraph? Yes — split answers are typical but not required. Shared-paragraph pairs need extra breaker scrutiny, not rejection.

How do I handle US/UK spellings? Normalise silently when scanning (organize/organise, percent/per cent) but copy exactly when answering. Spelling variants never change verdicts; limits and forms do.

What if the whole paragraph is unknown vocabulary? Drop to structure: first sentence for topic, last for verdict, names and numbers as anchors. Answer location questions from anchors alone, skip opinion questions on that paragraph, and return only with spare time.

How do I revise topic sets so they stick? Interleave them: mix environment pairs with health pairs in each review, recalling the proposition before checking the pair. Blocked review (one topic at a time) feels fluent and tests nothing; interleaved recall mirrors the test’s own mixing.

Should beginners start here or with foundations? Foundations first (skim/scan/detail in the companion R 01 article), then this engine. Paraphrase training without zone-finding wastes moves on unlocated text; zone-finding without paraphrase locates answers you cannot recognise. The sequence in the study plan above orders them accordingly — gears first, engine second, timed tests last — in that order, without skipping.

British or American spelling? Copy the passage exactly; this guide uses British English elsewhere.

The matching book

This article teaches the vocabulary-and-paraphrase engine behind IELTS ACADEMIC R 02: The Complete IELTS Reading Workbook (Academic) — paraphrase recognition, vocabulary-in-context inference, title choice and list selection — with the book’s drill bank, practice tests and explanatory keys as the training ground. The book adds the full 14-type bank, beginner’s-guide chapters and volumes this article samples. The matching edition is listed on staging here: product page; no price, format claim or availability is made here.

Continue your practice with IELTS ACADEMIC R 02: The Complete IELTS Reading Workbook (Academic)

Want more practice? This article introduces the key principles, but IELTS ACADEMIC R 02: The Complete IELTS Reading Workbook (Academic) provides additional practice, worked examples and structured preparation. View the book →

Sources and editorial note

Source read: Source-Books /IELTS/Reading/IELTS ACADEMIC R 02.epub (118 zip entries: vocabulary/paraphrasing chapters with exercises, 14-type drill bank, practice tests with explanatory keys, time-management and test-day chapters; full-chapter scan 2026-09-27, vocabulary and title chapters verified in full). Exercise patterns adapted, not reproduced; all passages and items here are newly written. No page numbers cited (EPUB, reflowable). Exam requirements should be checked against current official IELTS guidance. Published on staging 2026-09-27; staging product page verified live.