Most IELTS vocabulary study is aimed at the wrong target. Candidates memorise rare words from frequency lists, then sit a reading test that never asks them to produce a single word — it asks them to recognise that two phrasings mean the same thing. This lesson explains what reading vocabulary actually is, and installs the mining method that builds it from passages you have already done.
What the test actually pays for
Look at where reading marks are decided. A True/False/Not Given statement says purchased; the passage says acquired. A summary gap's frame says at first; the passage says initially. A correct MCQ option restates few researchers accept this as the theory has little support. Every one of those marks is a vocabulary event — but not a rare-word event. The words involved sit in the mid-frequency academic band: acquire, decline, initially, assess, considerable, subsequent.
The rule: reading marks come from recognising academic paraphrase families — clusters of everyday-plus-academic words that the test treats as interchangeable. Depth in 1,500 mid-frequency words beats a nodding acquaintance with 8,000.
That is why word lists fail. A list teaches you ubiquitous exists. The test needs you to see widespread, common, found everywhere, and ubiquitous as one unit, at reading speed, in four different grammatical costumes. Lists give breadth without connection; the test pays for connection.
Recognition beats production — so train recognition
For reading, you never need to spell these words, pronounce them, or use them in a sentence. You need three things, in order of value:
- Instant meaning — no translation pause.
- Family membership — which other words the test swaps it with.
- Form flexibility — recognising rely / reliance / reliable / reliably as one item, since the test shifts word class constantly (a transformation you will drill in paraphrase recognition).
This changes what "learning a word" means. Instead of word → definition, the unit you record is word → its swap set.
The mining method: your passages are the corpus
The highest-yield vocabulary source is not a list — it is the tests you have already answered, because every word there has proven it appears in real exams, in context, at test difficulty. The method:
- After scoring a passage, harvest — don't re-read. Go to each question you answered (right or wrong) and pull the paraphrase pair: the question's phrase and the passage's phrase. significant expansion ↔ grew considerably. These pairs are pre-certified test vocabulary.
- Add unknown words only if they blocked a mark. A word you didn't know in a sentence no question touched costs nothing; skip it. A word that made you miss the located sentence goes in.
- Record as families, not entries. One line per family:
decline — decrease, drop, fall, diminish, downturn (n/v). Add to existing lines as new members appear across tests; a family you have met in four different passages is a family you own. - Review by covering one side. See considerable → produce the family. Five minutes, twice a week. Recognition review is fast because recognition is the whole goal.
Ten minutes of mining per completed test yields 8–12 certified items. Across the twenty-plus tests of a normal preparation cycle, that is a few hundred families — chosen by the exam itself.
The glossary shortcut
You do not have to build every family from scratch. Our glossary collects the recurring academic vocabulary from the test library — the words that keep appearing across passages, with the senses the tests actually use. Use it two ways:
- After mining, look up the words you harvested and steal the family members you hadn't met yet.
- Before a timed test, skim ten entries as a warm-up; meeting three of them in the passage is the normal outcome, which tells you something about how concentrated this vocabulary really is.
What the glossary does not replace is the mining itself. A family you extracted from a question you personally missed is welded into memory in a way no browsed entry can match.
What to deliberately ignore
- Passage-topic jargon. Photosynthesis, cuneiform, thermohaline — the passage always defines or glosses the technical terms it needs. They are never the paraphrase pivot. Do not record them.
- C2-level ornaments. Words the test itself avoids can't cost you marks.
- Producing these words in essays. Reading vocabulary is a recognition asset; forcing harvested words into your writing is a different project with different rules.
The discipline of ignoring is what keeps the system alive. Vocabulary methods die of obesity — twenty minutes a day of recording everything, abandoned within two weeks. Mine only what the questions certified, and the habit costs ten minutes and survives.
Your drill
Twenty minutes, starting from a test you answer today or one you have already scored.
- Do (or reopen) Passage 1 and 2 of Reading 2026-02 Test 3. With the answers checked, harvest every paraphrase pair from the questions — aim for at least eight pairs written as family lines.
- Take the five families that feel least familiar and check them against the glossary; add any swap-members you find there.
- Tomorrow, before doing anything else, cover the right-hand side of your list and reproduce each family from its headword. Anything you can't reproduce gets a mark and reappears in two days.
- Repeat the harvest after your next test — Reading 2025-12 Test 1 is a good candidate — and append to existing families wherever possible rather than opening new lines.
Within two weeks the effect shows up somewhere unexpected: not in "knowing more words," but in location speed. You scan for funding and your eye now also trips on investment and financial support — the family fires as one unit, which is precisely the skill the next lesson trains to full depth.