CELPIP vocabulary: the right word depends on who you are writing to
Aman Batth · 10 min read · · Updated
Every CELPIP vocabulary page hands you idioms, connectors and synonym pairs. None of them answers the question the email task actually asks — how formal is this? The prompt names your recipient and never names the register, because judging it is part of the task.
Search for CELPIP vocabulary and you will be handed a list. Dozens of transition words, a long run of idioms with their meanings, pages of synonym pairs. The lists are usually accurate and they are not the problem.
The problem is that no list can answer the question the writing task actually puts in front of you, which is not what is a better word for good but who am I writing to, and how formal should this be.
We build the CELPIP practice scorer on Hilingo, so this is written from the marking side. The official performance standards are published by Paragon — read them at celpip.ca rather than taking any blog's word for the wording.
What the email prompt gives you, and what it withholds
Writing Task 1 is an email: 150–200 words, 27 minutes. The prompt gives you three things:
- a scenario — your neighbour's tree came down on your fence, your order arrived damaged, your team's shift pattern is changing
- a recipient, named in ordinary words: your building manager, your coworker, a friend
- a short list of bullet points saying what the email must contain
It does not tell you the register. That omission is deliberate, and it is the single most useful thing to understand about CELPIP vocabulary.
When we author an email item, the prompt carries an expected register — formal, semi-formal or informal — as a field on the question. It is used by the scorer and it is never sent to the student's screen, because the real task works the same way. Inferring the register from the recipient and the situation is part of what is being assessed. A word list cannot do that inference for you, and neither can we.
The dimension a misjudged idiom lands on is not vocabulary
CELPIP writing is rated on four dimensions: content and coherence, vocabulary, readability, and task fulfilment. Speaking uses the same four with listenability in place of readability. (CELPIP performance standards) If you want the levels those dimensions roll up into, that is how CELPIP is scored.
Here is the part that changes how you should study.
Register is not a vocabulary question. Register and bullet coverage are task fulfilment signals — that is how our own scorer weights them, and it follows the shape of the rubric. So when you drop it was a piece of cake into a formal complaint to a property manager, the cost does not appear where you expected it:
- Vocabulary gains nothing. An idiom used correctly is a correct phrase, not an impressive one.
- Task fulfilment takes the hit, because the email is no longer doing the job the scenario set.
- Content and coherence often takes a second hit, because a reader who has to work out why the tone shifted is a reader you have slowed down.
One phrase, two dimensions, and neither of them is the one the phrase was supposed to improve. That is the whole argument against studying CELPIP vocabulary as a list of upgrades.
And remember who checks it. Paragon states that CELPIP writing is scored by an AI-human hybrid system combining artificial intelligence with its human rating panel, and that all CELPIP writing scores are reviewed and verified by that rating team against the rating criteria. That verification step is a real strength of the test, and it is broader than what Pearson publishes for PTE Core, where a human expert sits alongside the AI on certain traits of certain tasks only. It also means your register is checked by someone who knows exactly how odd I hereby request sounds in a note to a friend.
Over-formal is also wrong
Most advice assumes formality is the safe direction: if you are not sure, be more formal. On CELPIP that is not safe, because informal prompts exist. Writing a stiff, official email to a friend is a register error in the same way that writing a chatty one to your building manager is.
| The prompt says you are writing to | Usually means | Fits | Reads wrong |
|---|---|---|---|
| your building manager, about a repair | Formal | I am writing to report…, at your earliest convenience | Hey, sort it out, a bit of a nightmare |
| a coworker you speak to every day | Semi-formal | Could you let me know whether…, I wanted to check | I hereby request, Dear Sir or Madam |
| a friend, about a weekend plan | Informal | Fancy coming along?, Let me know | I am writing to inform you, Furthermore |
Read the right-hand column again. Two of those three failures are caused by reaching upwards — by deploying exactly the formal connectors and set phrases that vocabulary lists are full of.
Furthermore is not a better word than and. It is a more formal word than and, which makes it the right choice in one of those three rows and the wrong choice in another.
The spell-checker changes which risk matters
The CELPIP writing screen has basic spell-check enabled, and our mock runner mirrors that — the writing textarea has spell-checking on, exactly as the real one does.
This inverts the usual vocabulary advice. If you are preparing for a test where spelling is marked cold, a long word you half-remember is dangerous because you might misspell it. On CELPIP that specific risk is largely handled for you.
What the spell-checker cannot catch is the word that is spelled perfectly and used slightly wrong. Comprise, alleviate, stipulate, mitigate — all of these will pass a spell-check while landing a fraction off the meaning you wanted. A human rater reads the sentence, not the spelling, and a word used a fraction off is read as a word you do not own.
So the risk on CELPIP is not can I spell it. It is can I use it without the rater noticing I have just learned it.
Speaking: eight tasks, eight different jobs
Speaking is where "learn more words" fails hardest, because the eight tasks do not want the same language.
Giving advice runs on modality — you could, it might be worth, if I were you. Dealing with a difficult situation runs on diplomatic framing, because you are choosing who to address and then having to be reasonable to them. Describing an unusual scene runs on precise concrete nouns, because the scorer is checking your description against what is actually in the image. Persuading runs on comparison.
A list of impressive adjectives serves none of those four well. The full task-by-task breakdown is in how CELPIP is scored; the point for vocabulary is that the useful unit of study is the phrase that does a job, not the word that sounds advanced.
There is also a hard constraint that written vocabulary advice always ignores: on speaking, the fourth dimension is listenability. A word you cannot say cleanly, at pace, with a countdown running, costs you on a dimension that your written practice never touches. If it slows you down or trips you up out loud, it is not yours yet — whatever your flashcard app says.
Listening and reading: you never produce a single word
Half of CELPIP-General asks you to recognise vocabulary, never to produce it. That is a different skill and a different study method, and synonym lists are close to useless for it — because you wrote the pairs, and the test picks a pair you did not think of.
The mechanic that makes this expensive: CELPIP listening audio plays once. No pause, no replay, no scrubbing back. A candidate who hears an unfamiliar word and spends the next few seconds deciding what it meant is not listening to the sentence that carries the next answer. Recognition has to be instant or it is not recognition.
So the drill is not "learn more words". It is: after a practice section, open the transcript and find the places where the recording used a word you were not expecting. On Hilingo, CELPIP listening transcripts are released after the attempt — never during it — and the answer review shows what you picked beside what was correct. Read the transcript for the swaps, not for comprehension. You already understood it on the second pass; that is not the information you need.
What to study instead of a list
- Register pairs, not synonym pairs. For each idea you actually use in emails — apologising, requesting, complaining, confirming, declining — hold one formal version and one informal version. I would appreciate it if you could and any chance you could. That is the choice the task is testing.
- Bullet-point verbs. The email lists three or four things to include, and each one is a task fulfilment signal. Practise turning a bullet into a sentence that visibly does what the bullet asked. No vocabulary buys back a bullet you skipped.
- Words you can say, not only spell. Say every new phrase out loud under a timer before you trust it in a speaking task.
- A swap log from your own transcripts. Every time a listening or reading item used a word where you expected another, write the pair down. Once the log runs to a page or two, you have a vocabulary list built from the test rather than from a blog, including this one.
The honest concession: vocabulary does matter. Range is one of four dimensions and a candidate who only has good, bad and very will be held down by it. But the range that scores is range you control, deployed at the formality the situation calls for — and that is a different project from memorising fifty idioms.
See which dimension your words are actually costing you
You cannot audit your own register. Your email sounded right to you when you wrote it, which is precisely why you sent it.
Hilingo gives you one full scored mock free, with no card. You sit a complete CELPIP-General or General LS test on the real timers, and every writing and speaking task comes back with an estimated level on each of the four dimensions separately — so you can see whether your task fulfilment is sitting below your vocabulary, which is what a register problem looks like from the scoring side. A model answer for the same prompt is released afterwards, and listening transcripts unlock with it.
One caveat we would rather say than hide: our scoring is our own engine, and nobody verifies it afterwards. Paragon says every real CELPIP writing score is reviewed and verified by its human rating panel; ours is an automated estimate of that judgement, not the judgement itself. Treat our levels as a direction and a diagnosis, not a prediction.
- How CELPIP is scored — the four dimensions, the eight speaking tasks, and what a 9 needs
- CELPIP score chart — CELPIP to CLB, IELTS and PTE Core
- PTE Core vs CELPIP — how much human review sits behind each score, and what changes
- IELTS speaking vocabulary — the same argument, run against the IELTS criteria
Frequently asked questions
How many words do I need to know for CELPIP?
There is no published number, and anyone quoting one has invented it. CELPIP rates vocabulary as one of four dimensions per task, on whether your word choice fits the situation and carries your meaning — not on the size of your vocabulary. A candidate with modest range used precisely scores above one with wide range used approximately.
Do idioms improve your CELPIP score?
Only when they fit the register. An idiom in an informal email or a relaxed speaking task is natural language and reads as such. The same idiom in a formal complaint is a task fulfilment problem, not a vocabulary bonus, because the email stops doing the job the scenario set. Learn a small number you genuinely use, and learn where they belong.
Is CELPIP writing formal or informal?
It depends on the prompt, and the prompt will not tell you. Task 1 names your recipient — your building manager, a coworker, a friend — and you infer the register from that and from why you are writing. Both directions are errors: an over-formal email to a friend loses marks in the same way a casual one to an official does.
Is there a spell checker in CELPIP writing?
Yes, the writing screen has basic spell-check. That removes most of the risk of using a longer word, but it cannot catch a correctly spelled word used with the wrong meaning — which is the failure a human rater notices immediately. Confirm current test-day conditions on celpip.ca rather than trusting any blog, this one included.
Do I need Canadian vocabulary for CELPIP?
CELPIP uses Canadian English and Canadian settings — rental situations, workplaces, community services — so the preparation that pays is familiarity with those everyday contexts, not a list of Canadianisms. You are far more likely to lose a mark on register than on a regional word.
How can I improve my CELPIP vocabulary in a month?
Stop collecting and start sorting. Take the words you already use, group them by formality, and practise the formal and informal version of the same intention. Then sit timed emails and speaking tasks and check the task fulfilment dimension rather than the vocabulary one — in a month, register is the fixable part and range is not.