“A man is going to lose his job because he looks unhappy.”
That sentence may sound imaginative, but a facial expression in a picture does not show the person's employment history or future. Scene tasks reward useful description and evidence-linked prediction, not an invented backstory.
CELPIP Speaking Task 3 is Describing a Scene. Task 4 is Making Predictions from the same illustration. The official 2026 Speaking study pack gives 30 seconds of preparation and 60 seconds of speaking for each task. It advises speakers in Task 3 to begin generally and then focus on selected details; for Task 4, it says predictions should relate to what people are visibly doing.
The lexical challenge is a boundary: Task 3 answers what is visible now; Task 4 answers what may happen next. This guide uses salient, seemingly, infer, plausible, and tentative to keep that boundary clear. Continue with the CELPIP vocabulary practice hub or compare exam paths on the Exam English hub.
Observation and prediction are different claims
| Evidence level | Suitable language | Example |
|---|---|---|
| directly visible | is holding, stands beside, in the foreground | “A customer is holding a receipt beside the counter.” |
| appearance-based | appears to, seems to, seemingly | “The cashier appears to be checking the receipt.” |
| evidence-linked future | may, is likely to, a plausible next step | “The cashier may process a return.” |
| unsupported backstory | avoid certainty or invented causes | Not: “The cashier sold a defective product yesterday.” |
An illustration can support an action, position, object, and some cautious descriptions of appearance. It rarely establishes private motives, relationships, earlier events, or exact future outcomes.
Task 3: build a useful map for the listener
The official prompt asks you to describe the illustration to someone who cannot see it. Start with an overview:
“This is a busy outdoor market with several customers moving between food stalls.”
The overview supplies setting and activity. Then organize details spatially rather than jumping randomly:
- in the foreground/background;
- on the left/right;
- near the centre;
- beside, behind, across from, between;
- closer to/farther from the entrance.
Choose a path—left to right, foreground to background, or central action outward—and keep it. The listener should be able to reconstruct the scene.
Select salient details
Salient describes a point or feature that stands out as relevant from a larger set. It is not merely a formal synonym for important.
“The most salient feature is a cyclist who has stopped in the middle of the path while two pedestrians move around him.”
The sentence selects one detail worth foregrounding. If you call every chair, cloud, and bag salient, the word loses its purpose. In spoken practice, the detail that stands out may sound more natural than repeating salient.
Prioritize details that create relationships:
- a person interacting with another person;
- an object affecting an action;
- a contrast between two parts of the scene;
- a visible cue that can later support a Task 4 prediction.
“There are nine people” may be accurate, but “Several people are waiting while one employee serves a customer” creates a more useful picture.
Describe actions with precise ordinary verbs
Replace generic doing, going, and having when the image supports a clearer verb:
- carrying, reaching for, pointing to, leaning against;
- queuing, browsing, loading, unloading, arranging;
- approaching, crossing, entering, leaving;
- speaking to, handing something to, looking toward.
Use the present continuous for an action visibly in progress: A child is reaching for a balloon. Use the simple present for stable relationships or scene facts: A bench stands beside the path.
Do not force specificity the image cannot support. If the object could be a folder or a menu, say a rectangular booklet or what appears to be a menu rather than asserting a hidden identity.
Use “seemingly” within its boundary
Seemingly means in a way that appears true from available signs while leaving open that fuller information could change it.
“A seemingly unattended suitcase is standing beside the bench.”
This reports appearance; it does not prove nobody owns the suitcase or imply deception. Natural forms include seemingly empty, seemingly quiet, seemingly minor, and seemingly at random.
Because a picture is already limited evidence, do not hedge every noun. A woman appears to be holding what seems to be a seemingly heavy bag is difficult to follow. Qualify only the uncertain element: A woman is holding a bag that appears heavy.
Keep feelings cautious
Official study materials mention describing people's feelings, but an expression is still visual evidence with limits. Prefer:
- He appears concerned;
- She is smiling and seems pleased;
- Their posture suggests they may be waiting impatiently.
Avoid diagnoses or histories: He is depressed because his marriage failed. The image cannot establish either claim.
Understanding is only the first step.
Lyra Practice helps you retrieve and use high-value workplace expressions in realistic situations until they feel natural.
Start a practice session →Task 4: turn a cue into a plausible next action
Task 4 uses the same illustration and asks what you think will happen next. Build each prediction from a visible cue:
“A server is carrying two plates toward a table, so she will probably deliver them to the waiting customers.”
The reasoning chain is cue → present direction → next action. Useful forms include:
- is about to + verb for an action that appears very near;
- will probably + verb for a reasonably likely forecast;
- may/might + verb for a possible outcome;
- a plausible next step is that... for a reasoned scenario;
- if X continues, Y may... for a conditional prediction.
These are not interchangeable decorations. Choose a strength that fits the cue.
Plausible is not probable
Plausible means reasonably possible or apparently valid on available facts. It does not mean proved or more likely than not.
“Because the cyclist is examining a loose chain, one plausible next step is that he will move the bicycle off the path to repair it.”
The cue makes the scenario reasonable. Another plausible outcome might be that he asks someone for help. If one outcome is better supported, call it more likely and explain why.
Do not stack uncertainty: perhaps a possible plausible prediction might be. One marker plus evidence is clearer.
Scale words require the same discipline. Substantial means large or important in amount, degree, or extent. A visibly long queue may support a substantial queue only when its size genuinely stands out; the word does not tell you why the queue formed or what will happen to it.
Infer from evidence; do not invent
To infer is to reach a conclusion from evidence. A viewer infers; a detail implies or suggests.
“From the open umbrella and wet pavement, we can infer that it has been raining.”
Even a reasonable inference is not guaranteed. Cannot safely infer is useful when a tempting claim exceeds the picture:
“We cannot safely infer that the two people are relatives simply because they are standing together.”
In Task 4, you do not need to announce the verb infer. Use the reasoning: name the cue, then make a calibrated forecast.
Tentative predictions still need content
A tentative prediction is held open to revision. The word marks the status of a conclusion, not a speaker's personality.
“My tentative prediction is that the queue will grow because three more customers are approaching the only open counter.”
This is a clear forecast with a reason. It differs from I am tentative to say, which is not the normal pattern. Use I am hesitant to say for personal reluctance.
In 60 seconds, saying my tentative prediction repeatedly sounds heavy. Alternatives include may, might, appears likely, and probably. Calibration should make the claim precise, not ceremonial.
Worked pair: one scene, two responses
Imagine an illustration of a public park. In the foreground, a child is reaching toward a kite caught in a low tree. An adult beside him holds a folded ladder. Two cyclists are approaching along the path. In the background, dark clouds are gathering while a family packs food into a basket.
Task 3 response
“This is a busy public park shortly before the weather changes. In the foreground, a child is pointing toward a red kite that is caught in a tree. An adult beside him is holding a folded ladder and looking up at the branches. The most salient safety detail is that two cyclists are approaching along the same narrow path, so the adult has not opened the ladder yet. Farther back, a family is placing containers into a picnic basket. One person is folding a blanket while another is watching the dark clouds. On the right, several empty benches stand near a playground. Overall, people seem to be finishing their activities rather than settling in.”
This response describes. The last sentence is a cautious synthesis from several visible actions, not a hidden story.
Task 4 response
“The adult will probably wait for the cyclists to pass before opening the ladder. Because the kite is caught in a low branch, it is plausible that he will then try to retrieve it while the child stands back. The family in the background appears ready to leave, so they may carry the basket toward the parking area. If the dark clouds continue to gather, other visitors are likely to move toward shelter as well. I cannot safely predict that it will rain, but the umbrellas and packed picnic suggest that people are preparing for that possibility.”
This response predicts. Each forecast follows from a cue. It does not invent names, motives, or outcomes beyond the scene.
Common mistakes across Tasks 3 and 4
- Listing without structure: organize details spatially and connect actions.
- Predicting during the entire description: keep Task 3 centred on what is happening now.
- Repeating the description in Task 4: use visible details as evidence for what happens next.
- Salient equals everything: select details that stand out from the larger scene.
- Seemingly guarantees reversal: it marks appearance, not deception.
- Plausible equals certain: a reasonable scenario remains open.
- Infer without evidence: name or imply the cue.
- Emotion as biography: a facial expression can support a cautious impression, not a life history.
- Future certainty from weak cues: will can fit a strongly directed action, but may/might may better reflect ambiguity.
A preparation method for the paired tasks
For Task 3, scan the image in three passes:
- Overview: setting, general activity, atmosphere supported by visible facts.
- Zones: foreground, centre, background; left and right.
- Links: who interacts, which object affects which action, and what stands out.
For Task 4, make three cue arrows:
- person reaching for item → may pick it up;
- vehicle approaching crossing → pedestrian likely to wait;
- people packing + dark clouds → may leave or seek shelter.
Aim for three developed predictions rather than ten unsupported guesses. Describe the cue before or inside each prediction so the listener can follow the logic.
Quick check
1. A customer holds cash toward a cashier. Which prediction is best supported?
A. “The customer may pay for the purchase.” B. “The customer will resign tomorrow.” C. “The cashier secretly changed the price yesterday.”
2. Which sentence uses seemingly accurately?
A. “The seemingly empty chair may be reserved.” B. “Seemingly proves nobody owns the chair.” C. “The chair is a seemingly.”
3. Which conclusion cannot safely be inferred from two people standing side by side?
A. They are near each other. B. They are members of the same family. C. Both appear in the scene.
Answers: 1-A, 2-A, 3-B. The cash supports a payment prediction. Seemingly qualifies appearance. Physical proximity does not establish a relationship.
Practice prompt
Choose any crowded illustration. Record two separate 60-second responses:
- Describe the setting, three spatially organized actions, and one salient feature.
- Make three predictions. For each, state the visible cue and choose may, probably, is about to, or a plausible next step deliberately.
Transcribe both recordings. Circle any biography, motive, cause, or future event the picture does not support. Rewrite those claims as observations, calibrated inferences, or evidence-linked predictions.
To practise precise description and prediction vocabulary, try Lyra Practice.
Sources
- CELPIP Test Format and Scoring — CELPIP
- CELPIP Speaking Pro Target 9+ Study Pack 2026 — Prometric
- CELPIP Results and Performance Standards — CELPIP
- CELPIP Speaking Pro — CELPIP
Related
- IELTS Speaking Part 3: Compare, Predict, and Speculate
- Cambridge C1 Speaking Part 2: Compare and Speculate
- PTE Describe Image Vocabulary for Maps, Processes, and Photos
- TOEIC Part 1 Photographs Vocabulary
Lyra Practice is not affiliated with or endorsed by CELPIP or Prometric.