Listening Practice Test 1
Start here. A community centre desk in Dartmouth, a garden centre in May, the ferry out to Gannet Island, a Halifax library dropping its late fees, and a talk on how the city clears its roads.
Free CELPIP listening practice for the 2026 format. Six complete listening mock tests, each with all 6 parts and 38 scored questions, built on 174 audio clips that start on their own and play once, exactly like the real test.
Six full-length CELPIP listening mock tests, all free and all live. Every one carries the same 6 parts and 38 scored questions, plus an unscored practice task, and they are graded Easy to Hard so you can start where you are and work up. Between them they run on 174 original audio clips.
Start here. A community centre desk in Dartmouth, a garden centre in May, the ferry out to Gannet Island, a Halifax library dropping its late fees, and a talk on how the city clears its roads.
A sticking back door at a hardware store, a curling club moving its league night, a transit counter on the move, a food bank deciding what to do next, and a talk on winter road safety, all set in Winnipeg.
The first of the medium tests. A conference poster that never got printed, a museum exhibit on cutting ice out of the lake, a mobile veterinary clinic north of Kingston, and a limnologist on what is really feeding the algae.
Six clips from Regina and southern Saskatchewan. A garden killed by well water, a community radio schedule shake-up, farmers buying the rail line they were about to lose, and a hard look at what a rural broadband map actually measures.
The top of our medium band. A school fundraiser that loses its cider press the week it needs one, a hotel front desk, a backcountry trail that starts asking for a day pass, and a strata council split over short-term rentals.
Our hardest listening set, where speakers change their minds and the answer is rarely in one place. A warehouse pay mix-up, a condo board election, a night school deadline, and a land economist correcting her own report.
This is the most useful fact about the CELPIP listening format, and most guides skip it. The section is not one question type six times over. It is two completely different question engines, and they split the 38 questions exactly down the middle, 19 each. Walking into Part 4 expecting a voice to read you the question is how candidates lose the first minute of a screen they only get one clock for.
Both halves are worth 19 of the 38 marks, so neither is optional. Our listening tests run both engines exactly as described.
The 2026 CELPIP Listening section keeps its 6 parts, its 38 scored questions and CLB 1 to 12 scoring. Here is the whole format on one screen so nothing on test day feels new.
Because listening runs first, it is the section you sit before you have settled in. That alone is a reason to rehearse it.
The CELPIP Listening section in the 2026 format has six parts carrying 38 scored questions, and it runs about 47 to 55 minutes. It is the first section on test day, which matters more than it sounds: you sit it before you have warmed up, and it is the one section where a slow start cannot be recovered later in the part.
The rule that shapes everything else is that the audio plays once. It starts on its own, you cannot pause it, and there is no progress bar to tell you how much is left. There is no version of this section where you go back for a second listen, so the skill being measured is not really comprehension in the abstract. It is comprehension the first time, at speed, while writing things down.
The six parts step up in difficulty as you go. Listening to Problem Solving opens with two strangers working through a problem, split into three chunks of conversation with questions after each. Listening to a Daily Life Conversation is the mirror image: two people who do know each other, in one unbroken clip. Listening for Information puts one speaker who knows something opposite one who wants to know it. Listening to a News Item switches to a formal single-voice report. Listening to a Discussion brings in three speakers who sometimes disagree. Listening for Viewpoints ends with a three-minute prepared talk that uses hard vocabulary on purpose.
Listening is marked, not judged. Every question has one correct answer out of four, it is scored by computer against a key, and nothing is deducted for a wrong answer, so a blank is a wasted 25 percent. Your raw score out of 38 converts to a level on the CLB 1 to 12 scale, and most competitive permanent residence applicants want CLB 9 or better, which is 33 out of 38. Paragon also mentions something worth taking seriously: there may be one part that is not scored, and you are never told which one, so there is no part you can safely coast through.
All 6 of our CELPIP listening mock tests cover the full six parts with 174 original audio clips, and they are scored the instant you finish, against the answer key, with the reasoning shown on every question. There is no waiting and no evaluation queue, because with a fixed key there is nothing to wait for.
Each part has its own speakers, its own length and its own question count, and they get harder as you go. Here is what each one actually asks of you.
Two speakers, a man and a woman who do not know each other, work through a problem one of them has. The conversation is cut into three sections of roughly a minute, and two or three questions follow each section. Informal, polite, and the gentlest listening on the test.
One clip of about two minutes, again a man and a woman, but this time they do know each other. That single difference is the whole distinction from Part 1, and you can hear it: people who know each other leave things unsaid, and the questions live in the gaps.
Two speakers over about two and a half minutes, where one has expertise the other wants: a booking, a procedure, a decision to be explained. Slightly more formal, and dense with the exact facts, figures and dates the questions ask for. The last part where you hear the question read to you.
One voice, about ninety seconds, formal and factual, the way a radio or TV report sounds. The first part where you read the questions instead of hearing them. Watch the names: news writing gives a person's full name once, then uses the last name for the rest of the item.
Three speakers, and the only part the real test delivers as video. They trade facts, opinions and feelings, and they will sometimes disagree with each other. Eight questions makes it the joint-biggest part, and keeping three people straight in your notes is most of the work.
One speaker, about three minutes, the longest clip on the test. A formal talk written in advance, dealing in concepts and analysis, and it uses low-frequency specialized vocabulary on purpose. This part is meant to be hard. Ours does not soften it.
Nine CELPIP listening tips, every one of them tied to a rule the format actually enforces. Skip the generic advice about listening to podcasts. These are the things that change marks on a section where the audio plays once.
The clip starts on its own, and the first ten seconds are the ones you cannot afford to spend ruling columns. Draw your layout during the instruction and introduction screens: two columns for a two-speaker part, three for Part 5. By the time anybody speaks, you should only be filling in.
Every clip opens with one line telling you the setting and who is talking. It is free context and it is over in seconds. Read it properly: knowing you are about to hear a customer and a clerk, rather than two friends, changes which details you expect to matter.
The questions arrive in roughly the order their answers appeared in the audio. That is a gift: it means notes taken top to bottom can be read top to bottom, and when a question lands you already know roughly where on the page to look. Do not scatter your notes around the sheet.
News items and prepared talks introduce a person with a full name once, then refer to them by last name for the rest of the clip. If you noted "Sarah" and the question asks about Ms. Whitfield, you have lost a mark you had earned. Catch the full name the first time, and note the surname.
The correct option is never a quote. It always paraphrases what you heard, so an option that repeats the audio word for word is more likely to be a trap than a key. Match the meaning, not the wording. If you hear that someone struggled to stay awake in a lecture, the answer will say they almost fell asleep in class.
Several questions ask how a speaker feels, and speakers almost never say. Nobody announces "I am frustrated". They say the bus is late again, and the attitude is carried by the tone and the word they lean on. Note the emotion as you hear it, not afterwards, because the tone is gone the moment the clip is.
Nothing is deducted for a wrong answer, so a blank is strictly worse than a guess. With four options, a blind guess is worth a quarter of a mark on average and a guess informed by half the clip is worth much more. In Parts 1 to 3 that 30-second clock is hard: put something down before it runs out.
Paragon says there may be one unscored part and that you are not told which one. So there is no part you can coast through and no way to spot the freebie. The practical reading of that rule is simple: it does not exist for you, so play all six parts straight.
The most expensive habit on this section is chasing a word you did not catch. While you replay it in your head, the speaker has moved on and you are now missing the sentence that held the next answer. Focus on what you understood, not on what you did not. One lost detail costs one mark; freezing costs three.
Most practice sites tell you their material is identical to the real test. Ours is not, in two specific ways, and you deserve to know which ways before you rely on it rather than after.
Six parts, 38 scored questions in an 8, 5, 6, 5, 8, 6 split, plus the unscored practice task. Both question engines behave as the real ones do. Every clip autoplays, plays once, and offers no pause, no scrub and no progress bar, because a player you can rewind rehearses a different exam. Two-speaker clips cast one man and one woman, as the format requires.
The real test uses native speakers with Canadian accents. We checked every speech provider available to us and no Canadian text-to-speech voice exists on any of them, so our speakers are North American instead. We put the Canadian English where we could control it: every script uses Canadian spelling and vocabulary and is set in real Canadian places, from Halifax and Kingston to Winnipeg and Regina. Standard Canadian English sits very close to General American anyway, so what you lose is small, but we are not going to call it Canadian when it is not.
The discussion is the one part CELPIP delivers as video. Ours plays the audio with the three speakers shown as a labelled panel, left, centre and right. That is defensible for a reason Paragon states themselves: visual details are never tested. Nobody is asked what colour a shirt was. What the video genuinely gives you is three faces in fixed positions to label your notes against, and the labelled panel gives you exactly that, with three clearly distinct voices so attitude still lands. We would rather show you a real panel than fake a video.
A full test tells you your band. These walk one part at a time, with worked answers, so you can fix the part that is costing you marks.
Two strangers work through a problem, with the full worked answers.
Practise this part → Part 2Two people who already know each other, so the context is assumed not stated.
Practise this part → Part 3One speaker holds the prices, dates and conditions the other needs.
Practise this part → Part 4A single reader delivers one short report and you complete the sentences.
Practise this part → Part 5Three speakers disagree and you track who holds which view.
Practise this part → Part 6A prepared formal talk that weighs an issue and hides the speaker position.
Practise this part →Listening runs first on test day, but it is one of four. Strengthen all four skills to lift your overall CELPIP band. Every section is free and built for the 2026 format.
The section right after listening: correspondence, diagrams, information and viewpoints across 4 parts, on a per-part clock.
Start reading →Email and survey tasks in a timed editor, returned with expert feedback and a CLB band.
Start writing →Record all 8 speaking tasks on screen with real prep and response timers.
Start speaking →Learn by example, one question type at a time. No timer, no marking, just model answers and the reasoning behind them.
Open the course →Start with Test 1 if the format is new to you, or go straight to Test 6 if you want the hardest set we have. All six are free, all 38 questions, and all scored the moment you finish. No waiting, because there is nothing to wait for.