CELPIP Speaking Task 4 (Making Predictions): a band-9 template

Speaking Task 4, Making Predictions, shows you the same picture you just described in Task 3 and asks one question: what do you think will happen next? You get 30 seconds to prepare and 60 seconds to speak, and you record once. The task is graded on how well you talk about the future — so the single most expensive mistake is spending your 60 seconds describing the scene again. Every sentence should move the picture forward in time.

The Task 4 format at a glance

• You see a picture of an everyday scene with several people in it — normally the same image you described in Task 3.
• The prompt asks you to predict what will happen next: in the next few minutes, later that day, or as a result of what you can see.
• You get 30 seconds to prepare and 60 seconds to speak. There is one recording and no retake.
• There are no right answers. Nobody checks whether your predictions come true — raters score the language you use to make them.
• Task 4 is one of eight speaking tasks in a component that takes about 15 minutes, so the pace is fast and you need a plan before the timer starts.

What examiners score

Task 4 is rated on the same four criteria as every other CELPIP speaking task, each on the 1–12 scale. What changes is how each one plays out when the subject is the future.

• Content / Coherence — are your predictions relevant to what is visible, supported with a reason, and organised so a listener can follow who you are talking about?
• Vocabulary — do you have a range of future forms and probability words, or does every sentence start with 'I think they will'?
• Listenability — steady pace, natural intonation, and smooth links between one prediction and the next.
• Task Fulfillment — did you actually predict, rather than describe, and did you use most of the 60 seconds doing it?

Future forms that keep you predicting

The grammar of this task is small and learnable. You need three or four ways to talk about the future, used on purpose to signal how sure you are.

The core forms

• will / won't — for confident predictions: 'She'll pay for the apples and move on to the next stall.'
• be going to — when there is visible evidence right now: 'The boy is going to get his ice cream — he's already pulling his mother towards the stand.'
• might / may / could — for less certain guesses: 'The two friends might stay on the bench for a while longer.'
• be likely to / be about to — to grade your certainty: 'The vendor is likely to get busier', 'the man is about to stop at the fountain.'
• future continuous for the wider scene: 'In an hour, the stalls will be packing up and the square will be much quieter.'

Hedging and reasoning language

• Opinion frames: 'I think', 'I'd say', 'I imagine', 'my guess is', 'I suspect'.
• Probability words: 'probably', 'most likely', 'almost certainly', 'possibly', 'chances are'.
• Evidence links: 'because', 'since', 'judging by', 'given that', 'seeing as' — these tie the prediction to something in the picture.
• Sequencing: 'first', 'after that', 'once she's finished', 'by the end of the morning' — these give the listener a timeline.

Mix certainty levels deliberately. Saying 'will' for everything sounds flat; alternating a confident 'she's going to' with a hedged 'they might' shows range and sounds like a real person thinking out loud.

A template that fills 60 seconds

Run the same five moves on every picture. Each has a clear job, so you always know what the next sentence is and never stall.

1. Frame it (1 sentence): 'Looking at this scene, I think a few things are about to happen.'
2. Main prediction (2 sentences): pick the most active person or group, say what they will do next, and give the visible reason.
3. Two or three more predictions (3–4 sentences): one person or group each, each with a future form and a 'because'. Vary the certainty — one 'going to', one 'might', one 'will probably'.
4. Widen the lens (1–2 sentences): the weather, the crowd, or the place an hour from now.
5. Wrap up (1 sentence): 'Overall, I'd expect…'

Step 4 is your safety valve. If you reach 45 seconds and your people are used up, you can always predict the weather, the crowd, or the clean-up, and glide to the buzzer without repeating yourself.

Sample band-9 response

Context: the Task 3 picture was a busy outdoor market on a sunny morning — a woman holding an apple at a fruit stall, a small boy pulling her sleeve towards an ice-cream stand, a vendor serving a customer, two friends with coffee on a bench, and a man walking a dog near a fountain.

Looking at this scene, I think a few things are about to happen. First, the woman at the fruit stall will probably pay for her apples and then give in to her son — he's pulling her sleeve so hard that she's likely to walk him over to the ice-cream stand in the next minute or two. The vendor in the green apron is going to finish serving his customer, and since a queue is forming behind her, he'll most likely be busy all morning. The two friends on the bench look relaxed, so they might stay a while, but I suspect they'll wander over to the stalls once their coffee is finished. The man with the dog is about to stop at the fountain, because the dog is already pulling towards the water. And given how bright it is, the market will probably get even busier as the afternoon goes on. Overall, I'd expect a pleasant, uneventful rest of the morning — nothing dramatic, just people getting on with their day.

Notice what this answer does: five separate predictions, each anchored to a visible clue with 'because', 'since' or 'given'; five different future forms ('will probably', 'is going to', 'might', 'is about to', 'I'd expect'); a wider prediction about the weather and crowd to fill the back half; and a one-line close. Not a single sentence is pure description.

Common mistakes

• Describing instead of predicting. 'A woman is holding an apple' is Task 3. 'She's going to buy it' is Task 4. If a sentence has no future form in it, it is not earning credit here. Quick test: could you add 'next' or 'soon' and have it still make sense?
• Predictions with no reasons. 'The man will leave' sounds random; 'the man will leave because his dog is already pulling towards the exit' sounds observed.
• Only one future form. Ten sentences of 'I think he will' hurts Vocabulary and Listenability. Rotate will, going to, might, likely to, about to.
• Far-fetched predictions. Plausible and grounded beats dramatic. Raters reward coherence, not imagination, and wild guesses are hard to justify in the next sentence.
• Running dry at 35 seconds. If you only prepared one prediction, you will finish early and then repeat it. Plan three or four people and keep the weather-and-crowd prediction in reserve.
• Freezing because you don't 'know' what happens. There is nothing to know. Pick the most likely next step for each person and say it with a hedge.

Once the template is automatic, Task 4 becomes one of the most predictable minutes of the Speaking test. Frame it, predict for three or four people with reasons, widen the lens, and close. Keep every sentence in the future and the band follows.