CELPIP Test Format: Every Section, Task and Timing

CELPIP is a Canadian English proficiency test used for immigration and professional licensing. It is entirely computer-delivered: you sit at a workstation with a headset and a microphone, read every prompt on screen, and type or speak every answer. There is no examiner in the room and no paper. That format is deliberate — it keeps every candidate's experience identical and removes the interviewer nerves that live speaking tests introduce. The trade-off is that the computer, not a person, sets the pace, and it never waits for you.

All four skills — Listening, Reading, Writing and Speaking — are tested in one sitting, typically around three hours at the test centre once check-in and equipment setup are included. You cannot split them across days, and you cannot carry unused time from one part into the next. Treat that as a stamina test as much as a language test: the section that costs you marks is often the one you reach when your concentration is already spent, not the one you found hardest on paper.

1. Listening6 parts

   ~47–55 questions · audio plays once

2. Reading4 parts

   ~55–60 minutes · each part timed

3. Writing2 tasks

   53 minutes total · typed, no spell-check

4. Speaking8 tasks

   ~15–20 minutes · recorded, no examiner

≈ 3 hours, in one sitting — computer-delivered, with no examiner in the room.

Because the whole test is timed and one-directional, knowing the shape of it in advance is a genuine advantage. Every minute you spend on test day working out what a section is or how a task behaves is a minute stolen from actually answering. This guide walks through each section in test order, lists the official timings for all ten Writing and Speaking tasks, and explains how those responses become a score — so that nothing about the structure is new to you when it counts.

Listening

Listening is delivered as one continuous section through a headset, after a short unscored practice segment that lets you confirm your audio is working and get used to the on-screen format. There are six scored parts.

Across the six parts you answer somewhere in the neighbourhood of 47 to 55 multiple-choice questions, though the exact count varies slightly between test versions. Don't fixate on hitting a number.

The rhythm is the same for every part: a short preview window to read the upcoming questions and, for most parts, the answer choices — then the audio, which plays start to finish once, with no pause and no rewind. After it ends you get a little more time to finish selecting answers before the test moves on.

That one-pass constraint is the single most important thing to know about this section. It is why note-taking shorthand and prediction from the question preview are trained skills rather than optional extras. When you read the questions before the audio starts, you are effectively deciding in advance what to listen for — a date, a reason someone is annoyed, the one detail two speakers disagree on — so that the recording confirms an expectation instead of ambushing you cold. Jotting a quick abbreviation for a name or a number the instant you hear it also frees your working memory to keep following the conversation, rather than clinging to one detail and missing the next.

The six parts are also arranged to escalate. Early parts lean on a single clear exchange or one speaker; later parts, such as the discussion with three or more voices, ask you to track who said what and hold competing points at once. Knowing that curve is there means you can budget your sharpest attention for the parts that demand it instead of burning out early on the easier ones.

Reading

Reading is computer-delivered and runs roughly 55–60 minutes across four parts. Each part has its own on-screen countdown — you cannot borrow time from a later part, and once a part's timer ends it is locked.

These numbers vary slightly by test form, but the total lands around 38 questions. Part 4 gets more time per question because tracking several viewpoints takes longer to set up, not because the individual questions are harder.

You will meet four question types: standard four-option multiple choice; dropdown fill-in-the-blank (cloze), common in Part 1's reply email and Part 4's reader-comment response; paragraph matching, where Part 3 asks which labelled paragraph a statement belongs to; and vocabulary-in-context questions. There is no separate grammar or vocabulary section — every question is tied to a passage.

The variety of question types is worth understanding before test day, because each rewards a different reading habit. A dropdown cloze question is really testing whether the word you slot in fits both the grammar and the tone of the surrounding sentence, so you read outward from the gap. A paragraph-matching question is asking where an idea lives, not whether it is true, so you skim for the passage that owns the claim rather than reading every line in order. Vocabulary-in-context questions turn on the sentence around the word, not a dictionary definition — the same word can point in different directions depending on the passage. Naming the type as it appears keeps you from applying the wrong strategy to it.

Reading is scored purely on correct answers. There is no partial credit and no penalty for guessing, so a blank and a wrong answer are worth exactly the same: zero. Always fill something in. Because each part's timer locks independently, the practical rule that follows is to never let a single hard question swallow the minutes a whole part needs — mark your best guess, move on, and come back only if time survives.

Writing

Writing has two tasks, both typed, both timed independently — 53 minutes of writing time in total. Everything goes into a plain text box with no spell-check, so accuracy has to come from you rather than the software. There is no essay task and no handwriting.

The recipient in Task 1 matters as much as the situation — the same complaint written to a friend and to a company manager should sound noticeably different. Task 2 is not testing which option you pick; it is testing whether you can build a coherent argument with specific reasons and examples.

Tone is a large part of what these tasks measure, and it is easy to underrate. An email to a friend can be warm and loose; the same request to a company manager needs to be polite, structured, and a little more formal. Getting that register right — the greeting, the level of directness, the sign-off — is part of Task Fulfillment, not a stylistic flourish. In Task 2, the argument matters more than the position: a well-supported case for the option you privately disagree with will always outscore a thin defence of the one you prefer, because the raters are reading for reasoning, not opinion.

A workable split for a 27-minute task is 2 minutes planning, 20 minutes drafting, 5 minutes editing. For the 26-minute Task 2, shift to 2 / 19 / 5. The planning and editing minutes are the first things to disappear under pressure, so protect them deliberately. Two minutes of planning is what stops you from realising halfway through that you have forgotten to answer part of the prompt, and with no spell-check to lean on, the final five minutes are the only thing standing between your typos and the rater. Both feel skippable in the moment and rarely are.

Speaking

Speaking is entirely computer-delivered and takes about 15–20 minutes. You read each prompt on screen, get a short silent preparation period, then speak into a microphone. There is no examiner and no live conversation — every response is a monologue recorded for later scoring.

There are eight tasks. Each has its own preparation time followed by a fixed response time, and the microphone records automatically the moment the response timer starts. You cannot pause it, and when it ends, recording stops — even mid-sentence.

The tasks fall into three families: narrative tasks (2, 3, 8) where you describe something concrete, stance tasks (1, 5, 6, 7) where you take a position and support it, and one hybrid analytical task (4) where you reason from visual evidence.

Grouping the tasks this way is more than tidy labelling — it tells you how to spend the prep window. For a narrative task, the useful prep move is to fix the concrete details you will describe; for a stance task, it is to decide your position and line up a reason or two before the recording starts. Because the microphone starts on its own and stops on its own, an answer that trails off with ten seconds of silence reads very differently from one that keeps developing an idea to the end. That is why the target for every Speaking task is to speak the full time: unused seconds are marks you simply left on the table.

Prep time is silent — the real test does not allow you to talk yourself through it — and on screen you will typically see nothing but the prompt and a countdown. Practising that way matters more than it sounds, because rehearsing out loud builds a habit the real test will not let you use. The skill to train is planning in your head and then speaking cleanly on the first attempt, with no chance to restart.

Timings at a glance

Every Writing and Speaking task, with its official preparation and response time.

Only Tasks 5 and 6 get the longer 60-second prep. Only Tasks 1 and 7 get the longer 90-second response.

How CELPIP is scored

CELPIP reports a score on a 1–12 scale, separately for each of the four skills. You do not receive a raw "38 out of 50" count. The 1–12 CELPIP scale maps to the Canadian Language Benchmarks, which is why immigration and licensing bodies can read it — but the two are separate scales, and a CELPIP score should always be quoted as a CELPIP score.

For Listening and Reading, your pattern of correct answers is converted using Paragon Testing's conversion tables. Because scoring is banded rather than linear, a handful of extra correct answers near a band boundary can move your reported level. That is a good reason to fight for every recoverable point rather than writing off one missed question.

Writing and Speaking are rated by human raters against four criteria, each contributing roughly equally:

Because the four criteria contribute roughly equally, no single strength carries a response on its own. A candidate with a rich vocabulary who ignores half the prompt is capped by Task Fulfillment; a perfectly on-task answer written in the same three simple sentences throughout is capped by Vocabulary. The practical takeaway is that raising a Writing or Speaking score usually means finding your weakest criterion and lifting it, not polishing the one you are already good at.

Note the difference in the third criterion: Writing is scored on Readability, Speaking on Listenability. Grammar and pronunciation accuracy still matter in Speaking, but they are folded into Vocabulary and Listenability rather than scored on their own. Listenability is about whether someone can follow you without effort — your pace, stress and intonation — and explicitly not about having a particular accent. A clear speaker with a strong regional accent is not penalised for it; a rushed or monotone speaker who is hard to track is.

What surprises people

The audio plays once. Every Listening recording plays start to finish a single time. No pause, no replay, no second chance at a name or a number.

There is no transcript and no rewind. What you hold in working memory, plus whatever you wrote down, is all you have when you answer.

Speaking records automatically. There is no button to press when you are ready. The prep countdown ends, recording begins, and it stops when the timer stops — even mid-sentence.

Reading timers are per-part, not pooled. Finishing Part 1 three minutes early does not help you on Part 3. Once a part closes, it is locked.

Writing has no spell-check. It is a plain text box. Every typo is yours, which is why the five-minute edit pass at the end of each task is worth protecting.

What to do next

Start with format, not content. Learn the order of the six Listening parts and the four Reading parts well enough to name them without checking, so nothing on screen is a surprise on test day. Then run one timed task from each skill under real conditions — full timers, single pass, no dictionary — and see where the time actually goes.

From there, work the specifics: prediction and note-taking for Listening, per-part pacing for Reading, the 2 / 20 / 5 split and a reusable email structure for Writing, and prep-time shorthand for all eight Speaking tasks. Each of those has a dedicated lesson in the four skill courses, and the writing and speaking templates give you a shape to fill rather than a script to memorise.