CELPIP Speaking Task 4: How to Make Predictions That Score

You just spent 60 seconds describing a picture in Task 3. The screen moves on, and the same picture comes back with a new question: what do you think will happen next? Most test takers handle this badly in one of two ways. They describe the picture again, because that's what their brain is still doing. Or they invent a plot twist so dramatic that it has nothing to do with what's on screen. Both lose points. This post shows you the move in between: predictions that come out of what you can see, said in the language of probability.

Want to try it on a real clock? Take the free mock test, which runs the Speaking tasks in order → celpipuni.com/free-mock-test.

TL;DR

CELPIP Speaking Task 4, "Making Predictions," gives you 30 seconds to prepare and 60 seconds to speak. You look at the same illustration from Task 3 and say what you think will happen next. Raters use the same four categories as every Speaking task: Content/Coherence, Vocabulary, Listenability and Task Fulfillment. The structure that works: two or three predictions, each tied to something visible in the picture, each with a probable consequence, said with future forms and probability language ("is probably going to," "will likely," "might end up"). The main mistake is describing the scene a second time. That's Task 3, not Task 4.

What does CELPIP Speaking Task 4 ask you to do?

Paragon's official Speaking study pack describes Task 4 in one line: look at the same illustration from Task 3 and describe what you think will happen next. That's the whole task. There are no options to choose from and nobody to persuade. You get a picture you've already studied and a question about its future.

That reuse is the important detail. You already spent 30 seconds of prep and 60 seconds of speaking on this picture, so you know who's in it, where they are and what they're doing. Task 4 doesn't test whether you can see the picture. It tests whether you can move the scene forward in time and explain why.

The official strategies for Task 4 are short:

1. Use your imagination, but be realistic. Your predictions should relate to what the people in the picture are doing.
2. Ask yourself what the people might do next, and what they might feel as a result.
3. Look at the surrounding people and objects for more ideas.

Each of those points tells you the same thing: your predictions have to come out of the picture.

How long do you get for CELPIP Speaking Task 4?

30 seconds to prepare and 60 seconds to speak. When prep time ends, the screen moves on automatically, you hear "Start speaking now," and recording starts. There's no button to press and no way to ask for more time.

Source: Paragon's official CELPIP Speaking Pro study pack. Formats get revised, so confirm current timings on celpip.ca before test day.

Sixty seconds is short, and natural speech fills it faster than people expect. Two or three predictions, each with a reason and a consequence, usually fit. If you try five, none of them gets a reason.

How is Task 4 rated?

Task 4 is rated on the same four dimensions as every CELPIP Speaking task. Here's what each one means for a prediction task:

• Content/Coherence. How many ideas you have, how good they are, how you organize them, and whether you back them up with examples and details. For Task 4 that means predictions that follow logically from the picture, in a clear order, each one developed.
• Vocabulary. Word choice, using words and phrases suitably, range, and precision. For Task 4 that means verbs of consequence and probability ("is about to," "will probably end up," "is bound to") rather than "will" twelve times.
• Listenability. Rhythm, pronunciation and intonation; pauses, interjections and self-correction; grammar and sentence structure; variety of sentence structure. Future forms are where grammar slips show most here.
• Task Fulfillment. Relevance, completeness, tone and length. In Task 4, relevance is everything: a prediction that ignores the picture isn't answering the task, and stopping at 35 seconds is a length problem.

Your Speaking level comes from your performance across all eight tasks, so Task 4 alone won't make or break your score. But it's one of the easiest tasks to tidy up, because the fix is structural.

What structure works for Task 4?

Think of it as evidence → prediction → consequence, repeated two or three times.

1. Evidence: point at something specific in the picture. "The man on the ladder is reaching for the top shelf…"
2. Prediction: say what will happen, with probability language. "…so he's probably going to knock that box of paint cans over…"
3. Consequence: say what that leads to, for him or someone else. "…and the woman underneath will likely have to jump out of the way."

Then the next prediction, from a different part of the picture. Order them by time, near future first and later future last, so the listener can follow the story. Finish with one sentence about how the scene ends overall.

Using the 30 seconds of prep:

• Seconds 1–10: pick two or three "unstable" details. These are things in the middle of happening: someone reaching, a child running, a cloud moving in, a line getting longer. Unstable details are where predictions come from.
• Seconds 10–25: for each one, write a keyword for what happens next and a keyword for the consequence. "ladder → falls → paint everywhere."
• Seconds 25–30: decide your order and your closing line.

You described this picture in Task 3, so don't spend prep re-reading it. Spend it on the future.

Which phrases help with Task 4 predictions?

Mix your future forms. One "will" after another sounds flat, and range is part of both Vocabulary and Listenability.

• Near future, strong evidence: "is about to," "is going to," "is on the verge of"
• Likely but not certain: "will probably," "is likely to," "I'd expect," "chances are"
• Possible: "might," "could," "may well end up"
• Consequence links: "which means," "so," "as a result," "once that happens"
• Time sequencing: "in the next few seconds," "a bit later," "by the end of the afternoon"

These are tools, not a script. If every Task 4 answer you give opens with the same memorized sentence, raters will hear it.

What does a band-9 Task 4 answer sound like?

The prompt and answer below were written by CelpipUni as a teaching example. They are not an official CELPIP sample, and the level is our estimate.

Practice scene (described in words, since there's no image here): a community garden on a Saturday morning. In the foreground, a teenage boy is pulling a wheelbarrow full of soil across a narrow path. An older woman is kneeling beside a vegetable bed, planting seedlings, with her back to him. On the right, a little girl is chasing a dog that has a garden glove in its mouth. In the background, dark clouds are moving in over the fence, and a man is setting up a table with coffee and muffins.

Question: what do you think will happen next?

"I think this calm morning is about to get a lot busier.

First, the boy with the wheelbarrow. That path is really narrow, and he's looking at the dog, not where he's going. So chances are he'll clip the edge of the vegetable bed, and some soil will spill right next to the woman who's planting. She'll probably laugh it off, though, because she looks like she's done this a hundred times.

Meanwhile, the little girl is going to get her glove back eventually. The dog clearly thinks it's a game, so I'd expect a few laps around the garden first, and maybe a muddy paw print or two on someone's jeans.

And the bigger change is the weather. Those clouds behind the fence are moving fast, so in maybe twenty minutes it'll probably start raining. Everyone will end up crowded under the coffee table's umbrella, eating muffins, and the planting will have to wait until the afternoon."

That's about 160 words, which is a natural fit for 60 seconds at a comfortable pace.

Why we'd expect this to score around 9:

• Task Fulfillment: every prediction starts from something visible (the narrow path, the dog, the clouds) and stays realistic. Nothing comes out of nowhere.
• Content/Coherence: three predictions, ordered from the next few seconds to twenty minutes from now, each with a consequence. The opening and closing lines frame it as one story.
• Vocabulary: "clip the edge," "laugh it off," "crowded under," "the planting will have to wait." Ordinary words, used precisely.
• Listenability: future forms vary ("is about to," "chances are he'll," "I'd expect," "it'll probably," "will end up"), and the sentence lengths vary too.

For contrast, a typical band-6 or band-7 answer to the same picture starts with: "In this picture there is a garden. There is a boy with a wheelbarrow and a woman planting. There is a dog and a girl…" That's Task 3 again. Thirty seconds go by before the first prediction.

What are the most common Task 4 mistakes?

Describing instead of predicting. This is the big one. You're still in Task 3 mode, so you list what's in the picture. Raters score Task 4 on whether you predicted. Your first sentence should already be about the future.

Predictions with no link to the picture. "Then aliens land in the garden." It's creative, but the official guidance says to be realistic and base predictions on what the people are doing. If a rater can't see where your prediction came from, it reads as off-task.

Only "will." "He will fall. She will be angry. It will rain." Correct grammar, no range. Mix in "is about to," "is likely to," "might end up."

Too many predictions, no development. Five one-line predictions sound like a list. Two or three, each with a "so" or "which means," sound like reasoning.

Stopping early. Ending at 40 seconds with "that's all" costs you on length, which is part of Task Fulfillment. If you finish your plan with time left, add one more consequence: what happens after the rain, how the woman feels, what the boy does next time.

Your Task 4 checklist

• First sentence is already a prediction, not a description
• Two or three predictions, each tied to a specific visible detail
• Each prediction has a consequence ("so…," "which means…")
• At least three different future or probability forms
• Predictions ordered by time, near future to later
• One closing sentence about how the scene ends

Want to know if your predictions sound like reasoning or like a list? Get feedback on all four rating categories → celpipuni.com/evaluation.

Why Task 4 rewards curiosity more than imagination

People assume the prediction task rewards the most creative answer. It mostly rewards the most observant one. The best Task 4 answers we hear aren't the most dramatic. They're the ones where the speaker noticed that the path was narrow, or the boy wasn't looking, or the clouds were moving, and followed that detail to its logical next step.

That's why Task 4 practice starts in Task 3. When you describe a scene, train yourself to notice what's unstable: what's mid-motion, what's about to tip, who's looking the wrong way. Your Task 3 answer gets more vivid, and your Task 4 answer arrives half-built. You can drill this with any photo: describe it for 60 seconds, then predict for 60 seconds, without a break.

Tools and resources

• The official CELPIP site and study packs. These cover current timings and the official Performance Standards. Formats get revised, so confirm there.
• CELPIP Speaking Task 3: Describe a Scene Like a 9. Task 4 builds on your Task 3 picture: celpipuni.com/celpip-speaking-task-3
• Scene Snap. Turn any photo into Task 3 and Task 4 practice: celpipuni.com/games/scene-snap
• CELPIP Speaking templates. Structures for all eight tasks: celpipuni.com/templates

Frequently asked questions

How long is CELPIP Speaking Task 4?

You get 30 seconds to prepare and 60 seconds to speak. When prep ends, recording starts automatically. Confirm current timings on celpip.ca, since formats are revised from time to time.

Is the Task 4 picture the same as Task 3?

Yes. Paragon's official study pack describes Task 4 as looking at the same illustration from Task 3 and describing what you think will happen next. So you already know the scene, and your prep can go straight to predictions.

How many predictions should I make in Task 4?

Two or three, each developed with a reason from the picture and a consequence. That usually fills 60 seconds. Five undeveloped predictions sound like a list and leave no room for reasons.

Can I describe the picture in Task 4?

Only as evidence for a prediction, as in "He's reaching for the top shelf, so he's probably about to…" Describing the scene for its own sake repeats Task 3 and weakens Task Fulfillment.

What grammar should I use in Task 4?

Future forms and probability language: "is going to," "is about to," "will probably," "is likely to," "might," "could end up." Using a range of them is better than repeating "will."

Do my predictions have to be correct?

There's no correct answer, because nobody knows what happens next. Your predictions do need to be realistic and connected to what's visible in the picture.

What to do next

1. Today, find any busy photo (a street, a park, a kitchen). Describe it for 60 seconds, then predict for 60 seconds. Record both.
2. Listen back. Was your first Task 4 sentence a prediction or a description?
3. Count your future forms. If it's all "will," redo it with three different forms.
4. This week, run Tasks 3 and 4 back to back on a timer so the switch from describing to predicting becomes automatic.

Related reading

• CELPIP Speaking: The Complete Guide to a 9+
• CELPIP Speaking Task 3: Describe a Scene Like a 9
• CELPIP Speaking Task 5: Making Choices With 60 Seconds to Prepare
• CELPIP Speaking Scoring Criteria: The 4 Things Raters Score

The bottom line

Task 4 gives you 30 seconds and a picture you already know, then asks for 60 seconds about its future. Start predicting in your first sentence, tie every prediction to something visible, give each one a consequence, and vary your future forms. Most people's Task 4 problem isn't their English. It's that they never leave Task 3.

Practice Task 3 and Task 4 back to back under real timing → Take your free mock test.