CELPIP Listening Part 5 (Discussion): It Is a Video, and Three People Are Talking

What Part 5 actually is

You have been listening to audio for four parts. Then Part 5 starts and there is a video on the screen.

Part 5 — officially "Listening to a Discussion" — is the only part of the Listening test that uses a video clip, and the only part with three speakers. It runs about 1.5 to 2 minutes and carries eight questions, tied with Part 1 for the most in the section. Almost nothing written about CELPIP mentions the video, so people arrive expecting audio, get a screen, and burn the first fifteen seconds working out where to look.

That is the trivia. The consequence matters far more.

With three voices in play, the question is almost never "what was said" — it is "who said it". Attribution is a different skill from comprehension. You can understand every word of that discussion and still lose marks, because understanding a point and remembering whose mouth it came out of are two separate acts of memory. Only one of them is being tested here.

An honest note about the video

The official material confirms two things: Part 5 is a video clip, and it has three speakers. What it does not confirm is whether the speakers are visible on screen during that video.

So this page will not tell you. Plan for a clip you must attribute by voice alone, and treat any faces you get as a bonus. Pages that promise you three labelled talking heads are guessing.

The failure: voice-blur

Call it voice-blur. It sets in around the sixty-second mark of a three-person clip.

Up to then you are tracking people. After that your brain quietly downgrades the discussion into a single stream of arguments — all correctly understood, all detached from their owners. The content survives. The name tags fall off.

Then the options appear and every one of them sounds familiar, because every one of them is familiar. Three speakers give the test writer a free supply of wrong answers that are entirely accurate about the clip and entirely wrong about the person.

Two things make this worse in Part 5. The clip plays one time only, so there is no second pass to re-attach names. And you read the answer options rather than hearing them — the format page is explicit about this — so at the moment the trap arrives, your eyes are busy and your memory is on its own.

Worked example

(original AlmiWorld practice item)

Three colleagues are discussing a staff training day.

NADIA: Honestly, I'd put the whole thing online. Half the team commutes in from out of town for a day that's mostly slides, and if it's recorded, anyone who's sick that week can still catch up afterwards.

TERRENCE: I'm with you on the morning. The presentation part, sure — nobody needs to sit in a room for that. But the afternoon is the equipment workshop, and you can't learn that off a screen. You have to have your hands on the thing.

PRIYA: I'd keep all of it in person, including the slides. Every online session we've run, people log in and then quietly go answer their email. And the useful part isn't the training anyway — it's the twenty minutes at lunch where someone tells you how they solved a problem.

TERRENCE: That's fair about lunch. I'd still rather not drag people in for a slideshow.

Question: What is Terrence's position on the training day?

• A. The entire training should be delivered online, partly so it can be reviewed later.
• B. The presentation should be online, but the practical session should stay in person.
• C. The whole training should stay in person, because people pay less attention online.
• D. The training should be postponed until a larger room is available.

Answer: B.

A is wrong, and it is the option most people choose. Every word of it is true about the clip — someone really does argue for full online delivery with recordings for people who miss it. That someone is Nadia. Terrence opens with "I'm with you on the morning", and partial agreement is the single most common thing to over-hear as full agreement.

C is wrong the same way, from the other side. The attention argument is real and it is in the clip — it is Priya's. Terrence concedes one narrow piece of her case ("that's fair about lunch") and then immediately restates his own view. A concession is not a position.

D is wrong for the ordinary reason: nobody says it. Room size never comes up. It is there for the person who has lost the thread entirely and is picking whatever sounds sensible.

B is right because it is the only option that keeps Terrence's split intact — online for the slides, in person for the hands-on work. Notice that B is less forceful than A or C. Middle positions are easy to under-record in your notes, which is exactly why they get tested.

Three columns, and what they look like at ninety seconds

Everyone tells you to take notes. Nobody shows you the page. Paper and pen are provided at the test centre, so this costs you nothing but the setup.

Rule your page into three columns before the clip starts. Write S1, S2, S3, and overwrite them the moment names are used. Then put each point under its owner and nowhere else.

Here is that same discussion, as notes, at about the ninety-second mark:

NADIA (S1)          | TERRENCE (S2)        | PRIYA (S3)
--------------------+----------------------+---------------------
ALL online          | SPLIT:               | ALL in person
- commute           |   am online (slides) | - people ignore online
- recorded/catch up |   pm in person       |   -> email
                    |   (equipment, hands) | - lunch talk = real value
                    |                      |
                    | agrees w/ N on am    |
                    | concedes lunch to P  |

Three things that layout is doing for you:

• Fragments, not sentences. "commute", "recorded". Sentences cost you the next speaker.
• Agreement is an arrow, not a merge. "agrees w/ N on am" stays in Terrence's column. The moment you write a shared point in the middle of the page, you have built your own distractor.
• Concessions are marked as concessions. "concedes lunch to P" tells you later that this was a courtesy, not a change of mind.

The shape that works

1. Rule the three columns before the clip plays, not during. Setup time is comprehension time.
2. Catch the names in the first fifteen seconds. Discussions open with people addressing each other. If a name never lands, keep S1/S2/S3 and identify by voice.
3. Write in the column, never across it. One thought, one owner, every time.
4. Track relationships, not just claims. Agrees with / disagrees with / concedes to. The questions live in that layer.
5. Flag the middle position. One person usually holds a split or qualified view. Underline it.
6. When you read the options, check the name before the content. Content-match is what the distractors are built to win.

Timing, honestly

The clip is roughly 1.5 to 2 minutes for eight questions. The published "30 seconds to hear and answer each question" figure applies to Parts 1 to 3 only, and no per-part limit is published for Part 5 — so do not budget against a number you invented. What is confirmed: there is an on-screen timer, the test automatically moves forward to the next screen, and you cannot return to a previous part. Your notes are the only version of the discussion you will ever get back.

What actually costs you the mark

• Choosing an option because you recognise the argument, without checking whose argument it was.
• Hearing partial agreement — "I'm with you on the morning" — and filing it as full agreement.
• Treating a concession as a change of position.
• Writing full sentences for speaker one and nothing at all for speaker three.
• Merging two speakers into "they thought" the moment their views overlap.
• Losing the first fifteen seconds to surprise, which is where the names are.

Try one

[Try Listening Part 5 →] — one full task, marked, no card required.