Should high schools start later?
The model pre-empts the lazy-teenager assumption before it can be raised, then concedes the childcare gap and answers it on fairness rather than logistics.
Practise this questionFive untimed CELPIP Speaking Task 7 questions, each with a spoken model answer. Ninety seconds of argument with nobody to persuade and nothing on screen to lean on. Record yourself, reveal the model, then read how it takes a position and defends it.
Task 7 asks what you think. Should high schools start later in the morning? Should cities charge households for every bag of garbage? Should children's teams keep score? Should there be a fixed retirement age? Should one sick day need a doctor's note? One question, 30 seconds to prepare, 90 seconds to answer.
What makes it the hardest task on the test for most people is what it takes away. Task 5 gives you options on the screen and a person to argue with. Task 6 gives you a situation and a listener. Tasks 3 and 4 give you a picture. Task 7 gives you a sentence. There is nothing to describe, nobody to persuade, and no material anywhere except what you can generate. Ninety seconds is a long time to fill from an empty room.
Which is why the failure here is so distinctive. Candidates do not run out of English on Task 7, they run out of argument, and you can hear the exact moment it happens: around second fifty they start restating the position they took at second three, in slightly different words, twice more. Ninety seconds of that is a band 6 answer delivered in band 9 English.
One fact makes the whole task easier, and almost nobody uses it: you are not required to be sincere. The task asks what you think, but nothing checks, nothing follows you home, and no rater has an opinion about garbage fees. So do not choose the side you believe. Choose the side you can develop, which is often the side you find slightly annoying, because you have already argued with it in your head.
There is a narrow path between never mentioning the other side and giving it equal weight. Both edges cost marks. The path between them is worth more than any other single technique on this task.
Caps the answer. The rater is listening for whether you understood the issue or memorised a side, and an argument that never meets an objection cannot demonstrate the first one.
Reads as indecision and is penalised under Task Fulfillment. On the one hand, on the other hand, and no landing, is an answer to a question nobody asked. The prompt says what do you think, not what could be thought.
Fairness is a virtue in a conversation and a problem in a 90 second answer. If your concession is bigger than your case, you have argued the other side by accident.
"Now, I do understand the objection, and it is a real one. Buses run in shifts, and parents leave for work at eight, so a later start opens a childcare gap. That is not nothing. But we are asking teenagers to absorb the whole cost of an adult scheduling convenience, and I think we have the wrong group carrying it."
Read that last one again and notice how much work it does in four sentences. It proves you know the counter-argument. It refuses to strawman it, which is a judgement call a rater can hear. It supplies a concrete mechanism, buses and eight o'clock, rather than a vague nod. And then it answers on a completely different axis, fairness rather than logistics, instead of pretending the objection went away. That is the shape the published descriptors are pointing at when they give the top bands both concrete and abstract language.
There are three other endings that do the same job. Redesign the policy to answer the objection: recycling stays free, every household gets free tags, so the fee only bites the people throwing out six bags. Relocate the cause: the problem there is the adults, not the scoreboard, so deal with the adults. Resolve with a distinction: defend the right to retire at sixty-five absolutely, and keep it a right rather than a deadline. All three concede something real and none of them surrenders.
The 2026 CELPIP speaking format for the opinion task specifically.
Nobody is grading your politics. They are grading whether an argument has a shape and whether you can build one out loud.
One more technique worth stealing, because it is almost free: admit that your position is a changed one. "I think they should keep score, and I say that as somebody who used to believe the opposite." "What changed my mind is ..." It buys credibility instantly, it signals that the other side was taken seriously before you get anywhere near the concession, and it gives you a natural structure for the next thirty seconds. Nothing else on this page does that much for nine words.
And a warning about overstatement. It is tempting under a clock to make claims bigger, because big claims sound confident. They are also easier to dismiss. "Older workers are always better" is refuted by one counterexample; "age is a proxy for capability and it is a poor one" survives contact with reality. Hedging is not weakness on this test. It is scored.
Five ordinary public questions, five different argumentative shapes: a flat yes, a qualified yes, a changed mind, a refusal with a distinction and a one-word no with the diagnosis attached. Open one, record your case with no clock running, then reveal the spoken model answer. Your recording stays in your browser: it is never uploaded and nothing here is scored.
The model pre-empts the lazy-teenager assumption before it can be raised, then concedes the childcare gap and answers it on fairness rather than logistics.
Practise this questionA position with one condition announced up front and paid off at the end. Promise a condition and forget it and the rater hears the coherence problem.
Practise this questionThe easiest question here to answer with feelings and no argument. The model reframes the whole debate away from competition and onto adults pretending.
Practise this questionTwo reasons on genuinely separate axes, a workable alternative rather than only an objection, and a tension the model admits it cannot dissolve.
Practise this questionOne word of answer, and the diagnosis arrives with it: a rule that does the opposite of what it was written to do. Contagion and evidence then carry a reason each, so neither one restates the other.
Practise this questionIn the 30 seconds, ask which side you have an example for. That is your side. Sincerity is not scored and nobody will ever know, and arguing a position you find slightly irritating is often easier because you have already rehearsed the objections.
The first sentence writes itself. The counter-argument does not, and it is the piece that lifts the band. Spend the prep deciding what the other side's best point is and what you will say back to it.
Check that your reasons are actually different. Money and cost are the same reason. Money and dignity are two. If both your points would be answered by the same objection, you have one point.
"The town where my sister lives cut its landfill volume by roughly a third in the first year." Nobody is fact-checking, and the hedge roughly is doing real work. A specific claim you can attach a number to beats any amount of general truth.
"What I would not accept is a flat charge with no free allowance." Drawing a line at the end shows a considered position instead of a slogan, and it fills the last ten seconds with content rather than a recap.
Take any headline and give it 90 seconds out loud: position, two reasons, an example, a concession, a close. Do it three times a week and the shape becomes automatic, which is the only thing that survives a 30 second prep.
CELPIP Speaking Task 7 is Expressing Opinions. A short question asks what you think about an everyday issue, such as whether high schools should start later or whether children's sports teams should keep score, and you argue your position. You get 30 seconds to prepare and 90 seconds to speak. It is the seventh of the eight speaking tasks in the 2026 CELPIP format.
Task 5 gives you options on screen and a named person to persuade. Task 7 gives you neither. There is nobody to argue with, nothing to compare and no picture to lean on, which makes it 90 seconds of argument built out of nothing but your own head. It is also 30 seconds longer than Task 5.
You should mention the other side, but you must not give it equal weight. Acknowledging an objection and then answering it is a top band move. Presenting both sides evenly and never landing is indecision, and it is penalised under Task Fulfillment, because the task asks what you think rather than what could be thought.
No. The task never asks you to be sincere and nobody checks. Choose the side you have more to say about, not the side you happen to hold. A well argued position you do not believe scores higher than a sincere one you cannot develop, and swapping sides is a legitimate strategy rather than cheating.
Two or three, each properly developed, plus one concrete example. Ninety seconds is about 200 to 230 words. Five reasons at that length gives each of them 40 words, which is a list. Two reasons on genuinely different axes, such as capability and fairness, give the answer range as well as depth.
No. Task 7 is text only, a single question on the screen during your 30 seconds of preparation, and it stays there while you speak. There is nothing visual to describe and no options to weigh, which is exactly what makes this the hardest task to fill for candidates who have not planned.
It concedes. It says the case against is real, states the strongest counter-argument honestly, and then answers it rather than waving it away. That single move demonstrates that you understood the issue rather than memorised a side, and it is the clearest separator between a solid answer and a high one.
Reach for a specific example or a concession, never a repetition. Restating your position for the third time is the commonest filler and it earns nothing. One named town, one number, one person you have worked with, or one honest account of what the other side gets right will fill the same time and add content.
No. There is no timer and no evaluation anywhere in the practice course. You record yourself so that you can hear yourself, the recording stays in your browser and is never uploaded, and then you reveal a spoken model answer with its structure, key phrases and common mistakes.
Yes. Everything here follows the 2026 CELPIP speaking format: eight tasks, a timer that runs per task and resets on every advance, and Task 7 running 30 seconds of preparation followed by 90 seconds of recorded speech.
Task 8 closes the test with the strangest job on it: describing a thing that has no name down a phone line.
A circumlocution test with four obligatory beats, and candidates reliably drop the last one.
Learn Task 8 → 1The other 90 second task, where the argument is aimed at one named person instead of nobody.
Learn Task 1 →Every task, every question, every model answer, on one page with the topics listed in full.
See the samples →All 8 tasks back to back with the real prep and response clocks running.
See the speaking section →Five opinions, no clock, no score. Record 90 seconds, then find the moment you started repeating yourself. That moment is what the practice is for.