CELPIP Speaking Task 4: Making Predictions Strategy
CELPIP Speaking Task 4 is the only task that asks you to invent the future. Thirty seconds to prep, sixty seconds to speak, and a picture that's the same one you just described in Task 3. Same scene, different brief.
CELPIP is accepted by IRCC for Canadian permanent residency and citizenship applications. For most economic immigration programs, CLB 7 in Speaking is the floor; CLB 9 unlocks the maximum CRS language points. Task 4 contributes to the single Speaking band that decides where you land.
This is where most candidates lose marks twice. First by re-describing what they already described thirty seconds ago. Second by hedging so hard the predictions never commit to anything.
Below: the format, the quadrant prep method that beats the freeze, the certainty grammar that scores higher than tense-only future, the three-prediction structure that fills sixty seconds cleanly, and a real Celpify sample answer scored at CLB 9.
How Task 4 Works
CELPIP Speaking Task 4 gives you 30 seconds of prep and 60 seconds to speak. You predict what will happen next using the same picture from Task 3. No new image, no replay, one take.
Task 4 is fast. Thirty seconds of prep, sixty seconds of speaking, and no replay.
| Phase | Detail |
|---|---|
| Prep time | 30 seconds, no speaking, scratch pad available |
| Speaking time | 60 seconds, one take, no replay |
| Picture | The same one you saw in Task 3 |
| Question | What do you think will most probably happen next? |
| Target answer length | About 120 to 150 words at a natural conversational pace |
The picture matters. Task 3 asked you to describe the scene. Task 4 asks what comes after. Same visual, different cognitive job. Side by side:
| Element | Task 3 (describe) | Task 4 (predict) |
|---|---|---|
| Verb tense | Present continuous | Future and conditional |
| What's tested | Visual detail vocabulary | Inference and certainty grading |
| Sample sentence | There are five people sitting around a table. | The skeptical person will probably ask a hard question. |
| Common slip | Listing items, no scene flow | Re-describing the present scene |
The big mistake is treating Task 4 as Task 3 with a future verb bolted on. That earns a partial mark at best. The task wants prediction logic, not scene description in a different tense.
Same picture. Different question.
Your 30 Seconds of Prep
In the 30-second prep window, write four one-word anchors on your scratch pad: current action, key people, the pivot moment, and likely next steps. Don't try to draft full sentences. Four anchors beat half a written sentence every time.
Don't draft a full answer in the prep window. You can't write fast enough and you can't think your way to clean phrasing under that timer. Use the quadrant method instead.
Split the picture into four mental quadrants and ask one question in each, then write a one-word note on the scratch pad:
1. What's happening now?
The current action in the scene. Meeting, presentation, construction, debate, queue. One word.
2. Who's involved?
The two or three people most visible. Note each one's stance: confident, skeptical, distracted, hopeful.
3. What's the pivot?
What's about to change? A decision, a reaction, a result. This is the engine of all three predictions.
4. What comes after?
Two or three concrete next-step predictions, mentally ranked by certainty.
Elegant phrasing isn't the goal here. Your goal is four anchors, not a full draft. The moment you stop trying to think in full sentences and start writing one-word notes, the freeze breaks.
Future-Tense and Certainty Language
The CELPIP rubric rewards prediction language, not just future tense.
Future tense without certainty grading reads as a flat list. Using three certainty levels in one answer shows range, and that's exactly what the rubric rewards.
| Certainty | Phrases | Use it when |
|---|---|---|
| High | will / is going to / definitely will | Visual evidence makes the outcome very likely |
| Medium | will probably / is likely to / most likely will | Outcome is reasonable but not guaranteed |
| Low | might / could / may end up / there's a chance | Outcome depends on factors you can't see in the picture |
Three predictions all stamped with will feels mechanical. Mixing in a will definitely, a probably will, and a might demonstrates the inference skill the rubric is built to score.
One conditional sentence raises your grammar score
A single conditional sentence in your answer signals advanced grammar. The shape: If the presentation goes well, the team will probably vote in favour. If it doesn't, the meeting will run long. One conditional per Task 4 is plenty. Forcing in two starts to sound studied.
The Three-Prediction Structure
Structure your 60-second Task 4 answer as four moves: a one-sentence framing opener (8 to 10 sec), a high-certainty prediction grounded in a visible detail (12 to 15 sec), a medium-certainty prediction in tension with the first (12 to 15 sec), and a low-certainty prediction plus a tidy close (15 to 20 sec).
The shape below fills the time cleanly without rushing or stalling. Drill it on three practice prompts and the pacing becomes automatic.
Move 1. Frame the prediction (8 to 10 sec)
One sentence that ties what you see to what you'll predict. Try: Looking at this scene, I think a few things are about to happen. Short, anchored to the visual, sets up three.
Move 2. High-certainty prediction (12 to 15 sec)
The most obvious next event, grounded in a visible detail. The presenter looks confident and has well-prepared slides, so the meeting will probably go smoothly. Visible detail then prediction, in that order.
Move 3. Medium-certainty prediction (12 to 15 sec)
A second outcome, ideally in tension with the first. However, the person in the back looks skeptical, so there will likely be some challenging questions. Use a connector (however, although, but) to link the two.
Move 4. Low-certainty prediction and wrap (15 to 20 sec)
One uncertain outcome plus a tidy close. The team might split into smaller groups afterwards. Overall, I expect the meeting will end with some kind of decision, even if it takes a bit longer than planned.
Each prediction starts with a visible detail, like the presenter, the person in the back, or the documents on the table, and then commits to an outcome. Pairing a visible cue with a predicted outcome is the whole task.
60-Second Model Answer: The Office Meeting
One real Celpify prompt, one CLB 9 sample answer, full rubric annotation.
The scene (paraphrased)
A meeting room. Five people seated around a long table. A presenter at the front stands beside a screen showing a slide deck. Documents and notepads sit in front of each seat. One person at the back has their arms crossed and looks skeptical. The presenter looks confident and engaged.
Model answer (about 130 words, CLB 9)
Looking at this meeting scene, I can already see a few things that point to what's about to happen. The presenter at the front looks confident and has a well-prepared deck on the screen, so I think the meeting will probably go smoothly at the start. However, one person at the back has their arms crossed and looks skeptical, so there will likely be some challenging questions during the discussion. The documents in front of every seat suggest this is a decision-making meeting, so I predict the team will vote or reach a consensus by the end. After the meeting, people will probably split into smaller groups to talk through their reactions. If the presentation is persuasive enough, the skeptical person might change their mind. If it isn't persuasive enough, the team will need a follow-up meeting.
What this answer scores
- Vocabulary range: three certainty levels (probably, likely, might), connectors (however, so, if not), specific visual nouns (deck, arms crossed, documents).
- Grammar variety: simple future, going to future, conditional. One if-clause near the end.
- Task fulfilment: four distinct predictions, each tied to a visible detail rather than invented out of thin air.
- Fluency: short framing sentence, builds rhythm, lands a tidy two-part close.
The pattern is reusable. Look at the picture. Pick four visible cues. Stamp each cue with a different certainty level. Land the answer at around 130 words.
Mistakes That Cap Task 4 Scores
Three mistakes consistently cap CELPIP Task 4 scores: re-describing the Task 3 scene in future tense, staying in present tense throughout, and rephrasing one prediction three ways instead of building three distinct outcomes.
1. Re-describing the Task 3 scene
The most common slip. Half your answer turns into I can see five people, the presenter is wearing a suit, the room has a long table... That's Task 3 territory. Task 4 wants prediction, not description. Every sentence after the opener should contain a future or conditional verb.
2. Staying in present tense the whole time
Saying the skeptical person disagrees when you mean the skeptical person will probably disagree is a small word, big mark. The rubric scores future and conditional grammar. Skipping it caps you at about CLB 7.
3. One prediction stretched to sixty seconds
One idea rephrased three ways still counts as one prediction. The rubric wants distinct outcomes. If your three sentences could all be summarised as the meeting will go well, you've made one prediction. Build three by pulling cues from three different parts of the picture.
The thirty-second prep trap
Sources & further reading
Check the official Speaking format before you drill the prediction task.
- CELPIP-General Test FormatOfficial source for the test's section structure and timingOfficial CELPIPcelpip.ca
- CELPIP Free ResourcesOfficial CELPIP study webinars and resourcesOfficial CELPIPcelpip.ca
CELPIP Speaking Task 4: Common Questions
Quick answers to the questions test-takers ask before Task 4.
Thirty seconds, sixty seconds, one picture you've already seen. Use the quadrant method, stack three certainty levels, ground every prediction in a visible cue, and Task 4 stops being the task that catches everyone.