Speaking Tasks 3 & 4: Describing a Scene & Making Predictions
Task 3 asks you to describe a picture to someone who cannot see it. Scorers look for coverage of the main elements, position words, and rich descriptive detail.
No account needed for the demo. Practicing the full bank takes a free account.
Speaking Tasks 3 & 4 format: how the test plays out
Every panel below is a moment from the exam interface this site rebuilds. Read the flow once, then run the same screens yourself in the demo.
Task 3 shows a detailed picture. You get 30 seconds to scan it and pick four or five things worth describing.
Cover the main elements with position words and rich detail, for someone who cannot see the picture.
Task 4 brings the same picture back. You get 30 seconds to prepare a prediction.
Future forms and logical guesses tied to what you can see. Then the pair is done.
Both tasks run back to back on one picture: 30 seconds to prepare and 60 to speak, twice.
That is the whole flow. The demo runs these exact screens with a real question, timed like test day.
How Speaking Tasks 3 & 4 is scored
Trained raters score each Speaking task on four equally weighted dimensions: Content and Coherence, Vocabulary, Listenability, and Task Fulfillment. All eight tasks feed one Speaking result, reported as CELPIP levels 1 to 12 aligned with CLB.
Four equally weighted dimensions per task, including Listenability
All eight tasks count, there is no task you can safely throw away
Levels 1 to 12, aligned with CLB for immigration targets
Tasks 3 and 4 share one picture but are scored as two separate tasks, so a weak prediction cannot hide behind a strong description.
Practice here scores every recording on the same four dimensions plus grammar, with a word by word transcript and a CLB style estimate.
Submit and get the AI report
Your recording is scored on four dimensions plus grammar, with strengths and weaknesses on each one, grammar fixes pulled from your own words, a word by word transcript you can replay, and a CLB style estimate. Each note points at something specific to fix, so you know where the next point comes from.
Inside your Speaking score report
Every answer comes back as one report in two parts: a transcript of what you said, and the grading details that show where the score came from. Both are built from your own answer, and the improvement lives in the details.
First, replay your answer word by word
Click any word to hear yourself say it. A word the transcript got wrong is usually a word your pronunciation got wrong, and the thin stretches of your answer are easy to find and replay.
The same answer as readable text. Grammar points are marked right where they happened, and opening one shows the fix written on your own sentence.
Then, the grading details
Strengths and weaknesses on every dimension
Content and Coherence, Vocabulary, Listenability, and Task Fulfillment each get their own read: what worked, what fell short, and what to change. You can tell exactly which dimension is holding the score back.
See where you stand
Your result lands on a distribution of scores from every test taker on the site, and each dimension is set against the site average, so you know whether you are ahead, on pace, or behind on that skill.
A written review, sentence by sentence
Overall feedback reads your whole answer back: an overall assessment, sentence by sentence improvements with the why behind each change, vocabulary upgrades for words you actually used, and key sentences rewritten at three levels.
Grammar notes on your own sentences
Improvement points are marked right in the transcript. Open one and the correction is written on your sentence, so you fix the habit you actually have instead of a textbook example.
Speaking Tasks 3 & 4, answered
How do CELPIP Speaking Tasks 3 and 4 work together?
They run back to back on the same picture. Task 3 asks you to describe the scene, Task 4 asks you to predict what happens next, each with 30 seconds of prep and 60 of speaking.
Are Tasks 3 and 4 scored separately?
Yes. Each is scored as its own task on the same four dimensions, so both count toward the Speaking result independently.
What does the describing task reward?
Coverage of the main elements with position words and concrete detail, spoken for someone who cannot see the picture.
What does the prediction task reward?
Future forms and logical guesses tied to what is actually visible. A wrong prediction costs nothing if it is delivered clearly and grounded in the scene.
All CELPIP Speaking question types, covered
8 tasks · about 20 minutes on the mock
One scene, two tasks: describe what you see, then predict what happens next, paired like the real exam.
Ready to try Describing a Scene & Making Predictions?
Open the demo to see the exact exam interface, or create a free account to practice the full bank. Everything is free during pre-launch, no credit card required.
