CELPIP Speaking Task 4 is Making Predictions: 30 seconds of preparation, 60 seconds of recording, and the same illustration you saw in Task 3.
The official score report gives no Task 4 score. Put your practice effort into how this answer demonstrates the four official speaking dimensions, rather than hunting for per-task weightings that are not published.
Content / Coherence
The first question is whether your predictions connect to visual clues. CELPIP official advice is to be imaginative but realistic — predict what people will do next and what those actions might cause or how they might feel.
A complete content unit contains a visual clue, a future action, and a likely result. Seeing a child reaching after a rolling ball, you can predict that the ball will go into the road and the parent nearby may rush to stop them.
Do not simply describe again
Repeating the current scene does not fulfil a Making Predictions task. One sentence can bridge from the scene, but the body has to say what comes next.
Choosing three or four visible people and predicting for each is usually easier to keep relevant and organised than building one long story only loosely connected to the picture.
Vocabulary
This dimension looks at accurate, varied language for the future and for probability. Will, might, may, could, is likely to, and is probably going to are all available — but vary them for meaning, not for the sake of variety.
Action and result vocabulary should also be specific. Repeating I think and something will happen limits precision.
Listenability
Listenability covers whether the predictions come out at a natural pace, with pronunciation and sentence structures a listener can follow.
Conditional sentences connect clue to consequence efficiently: If the man does not notice the spilled drink, he may slip. Complex structures only help when you control them. Frequent self-correction, or long pauses spent scanning the picture for details, reduce the amount of scoreable content.
Task Fulfillment
This dimension checks that the predictions are relevant, sufficiently complete, natural in tone, and finished within 60 seconds.
A prediction does not have to be the only possible outcome, and raters do not judge whether the future was “guessed correctly.” What matters is whether the inference is reasonable and expressed in clear English. A story completely disconnected from the picture, or a continued description of the present, is far more likely to hurt Task Fulfillment.
How to practise
Record Task 4 immediately after a Task 3 description so you get used to the sequence the real test uses.
On replay, circle every future expression and check that it has a basis in the picture and a result. If most of your sentences are still in the present continuous, the task probably was not completed. The goal is coherent, specific, easy-to-follow predictive evidence — not counting how many outcomes you called correctly.
FAQs
Does my prediction need to be right? No. Reasonable and clearly expressed is what counts.
Can I re-describe the picture? Only one bridging sentence. The body must be about what happens next.
Which modal verbs should I use? Vary them by degree of probability, not at random.
Next, read how Task 3 is scored and how Task 5 is scored.