The trap is built into the order
Task 3 shows you an image and asks you to describe it. Task 4 shows you the same image and asks what will happen next.
You have just spent a minute describing. Your mouth is warmed up for describing. And the image is still in front of you. So most people describe it again, with a few future-tense words sprinkled in, and score badly on a task they understood perfectly.
Task 3 is the present. Task 4 is the future. Nothing you say in Task 4 should be visible in the picture.
What a weak answer sounds like
"The children are playing on the floor and the mother is working on her laptop. Maybe she will continue working. The dog is near the mess and maybe it will stay there. The father is cooking so probably he will finish the food."
Every sentence begins in the picture and barely leaves it. "He is cooking, so he will finish cooking" is not a prediction — it is a description with a future tense bolted on.
A prediction says something the picture does not show.
What a strong answer sounds like
*"I think the mother is about to notice the cereal on the floor, and the moment she does, the laptop is getting closed. She'll probably deal with the older child first, because he's the one who looks like he already knows he's in trouble.
The dog is going to get to that cereal before anyone cleans it up — it's already too close. So I'd guess someone ends up moving the dog out of the kitchen before they can start sweeping.
As for the father, I don't think he's noticed any of it yet. He's got his back to the room. I'd say breakfast is going to be about ten minutes later than he's planning."*
Nothing there is in the photograph. Every sentence is a consequence.
The shape that works
- Start with the most likely next event. The one thing that is clearly about to happen.
- Say what follows from it — a chain, not a list. "She notices → the laptop closes → she deals with the older one first."
- Give a reason for at least one prediction. "Because he looks like he already knows." This is what separates predicting from guessing.
- Predict something about a person who has not noticed anything. This almost always produces the most interesting sentence in the answer.
Commit
Hedging is what costs marks here, and it is the natural instinct. Nobody can know the future, so people protect themselves: "maybe… perhaps… it might be possible that…"
Six hedges in sixty seconds is not caution. It is a response that never says anything.
The fix is not to become certain. It is to use confident language about uncertain things, which is what fluent speakers actually do:
- "I think she's about to…"
- "That dog is getting to the cereal first."
- "He'll probably…"
- "I'd guess…"
- "By the time he turns around…"
Each of those is honest. None of them is timid.
Useful phrases for the future
- is about to / is going to / any second now
- by the time he turns around / once she sees it
- ends up doing / will have to
- my guess is / I'd say / knowing kids that age
Notice how many carry a sequence. Task 4 rewards a chain of events, and these are the words that build one.
What is not being marked
- Not whether you are right. Nobody knows what happened next. There is no photograph of the next minute.
- Not your accent.
- Not how many predictions you make. Three connected predictions beat eight disconnected ones.
And the rule underneath everything:
A response giving insufficient information to assess scores 0 or 1.
A minute of re-description can reach that. If a marker cannot find a prediction in your answer, you did not do the task.
One habit to practise
Do Task 3 and Task 4 on the same photo, back to back, exactly as the test does. Then check whether a single sentence appears in both. If one does, that is the sentence costing you the mark — and catching it in practice is the entire skill.
Try one
[Try Task 4: Making Predictions →] — one full task, marked, no card required.