The whole Speaking test under exam conditions. Start the clock, then record each part in one take — no stopping, no restarting, no reading the model first. Twelve minutes for the four parts, as in the real room. Only when all four recordings exist do you open the models and listen back. The gap between what you said and what the model says is the debrief.
Under exam conditions Speaking is timed as the real test is: record each part in one take, without stopping to think, then listen back against the model. There is nothing to hand in — the recording is the paper.
The examiner asks you and your partner some questions about yourselves. Record all five answers in one take, leaving a two-second pause between them as the examiner would. Two minutes is the whole part — about twenty seconds an answer. Two sentences, every time. Fact, then detail.
The debrief — read this after you have recorded. A strong Part 1 is not clever; it is complete. Every answer is two or three sentences long, every opinion has a because attached, and no answer stops on a full stop that leaves the examiner waiting.
“I usually cycle, which takes about fifteen minutes if the lights are kind to me. In winter I take the bus instead, because the cycle path isn't lit and my mother worries.”
“Next weekend I'm going to my cousin's — she's just moved to Porto, and I haven't seen the flat yet. We'll probably spend Saturday buying furniture and Sunday putting it together.”
“Winter, actually, which surprises people. I like the dark evenings, because everyone stays in and the house is full. Summer is nicer, but it's emptier.”
What usually went wrong in the one-take version. The first answer was fine and the fifth was one sentence — energy drops through the part, and the examiner hears it. The future question got a present tense (I go to my cousin's). The why in the last question was answered with because it's nice, which is not a reason. Listen to your recording for those three things specifically.
Three habits worth more than new vocabulary. Finish the sentence. Add a reason to every opinion. Slow down — speed hides the endings of your words, which is where your grammar lives. And if you did not understand a question, the sentence that saves you is Sorry, could you say that again? It costs nothing; silence costs a lot.
The examiner gives you a photograph and you talk about it on your own for about one minute. Then your partner does the same with a different photograph, and you are each asked one short question about your partner's picture. There are no pictures on this page, so the photographs are described in words below. Read your photograph once, start the recording, and talk for a full minute without stopping. Then answer the question about your partner's picture in the same take.
Your photograph — students working together. A classroom with the desks pushed into a square. Five students are standing around a large sheet of paper on the desks, drawing a plan of something with coloured pens. One girl is holding a ruler and pointing; a boy opposite her is laughing and shaking his head. Another boy is sitting on a desk at the edge, looking at his phone. A teacher is at the back of the room, leaning on the wall with her arms folded, watching but not joining in. On the whiteboard behind them someone has written “45 minutes left” and underlined it twice.
Your partner's photograph — travellers at a station. A big railway station with a glass roof, early in the morning. A family with three enormous suitcases is standing under the departures board; the mother is looking up at it, the father is looking at his phone, and a small boy is sitting on one of the suitcases eating a croissant. Behind them a man in a suit is running with a coffee in one hand. Two young women with backpacks are asleep against a pillar. A station worker in an orange jacket is pointing the way for an elderly couple.
The debrief — read this after you have recorded. First, look at the length of your recording. Under a minute for the picture is the commonest result of this part under exam conditions, and it is a shape problem, not a vocabulary problem: a list of objects runs out at fifteen seconds. A minute needs four moves.
“This photograph shows a classroom, but not a normal lesson — the desks have been pushed together into a square, and I think the students are doing a project, because they're drawing something big on a sheet of paper. In the middle a girl with a ruler is pointing at the plan, and the boy opposite her is laughing and shaking his head, so I think they disagree about something. On the right there's a boy sitting on a desk looking at his phone — he isn't really taking part. At the back the teacher is leaning on the wall with her arms folded. She isn't helping; I think she wants them to work it out themselves. On the board someone has written that there are forty-five minutes left, so they're probably a bit stressed. It looks like the kind of lesson I'd enjoy, actually, apart from the boy on the phone.”
The four moves, and where yours probably stopped. (1) Where and what — with a because. (2) Two or three people, present continuous, with position words: in the middle, opposite her, on the right, at the back. (3) A guess with a reason: so I think, probably. (4) The one you keep looking at, or a personal reaction. Most one-take recordings do (1) and half of (2), then stop. If yours stopped at thirty seconds, it stopped between move 2 and move 3 — and move 3 is the easiest one, because guessing needs no facts.
The partner question. Thirty seconds, not another minute: Yes, I do, because you can read or sleep, and nobody expects you to talk to them. On a coach I always feel sick. Answer, reason, one detail, stop. Talking over the end of this question eats your partner's time and the examiner notices.
Now you talk to your partner, not to the examiner. You are given a situation and some ideas, and you discuss them together and then agree. It is marked on the conversation — listening, responding, disagreeing politely, bringing the other person in — not on the decision. Record it in one take: with a real partner if you have one, or alone playing both parts. Two to three minutes, and reach the decision out loud before you stop.
A teenager from your country is visiting London for the first time and has just one free day. Here are some things they could do. Talk together about how interesting each of these would be for a first visit — and then decide which two they should do with their day.
A · The British Museum — the mummies, the Rosetta Stone, and it is free.
B · A boat trip on the Thames — from Westminster to Greenwich, an hour on the water.
C · A football stadium tour — the dressing rooms, the tunnel, the pitch.
D · A street market — food from every country, clothes, music, crowds.
E · A walk in one of the big parks — deer, a lake, and quiet in the middle of the city.
F · A musical in the West End — an evening show in a famous theatre; tickets from £25.
The debrief — read this after you have recorded. Listen to your recording and count three things: how many times you asked your partner a question, how many times you started a sentence from something they said, and whether the decision was said out loud. Those three counts are most of the mark in this part.
“Shall we start with the museum? I think it's the obvious one for a first visit, because it's free and it's full of things you've seen in books. What do you think?”
“I agree it's famous, but you could spend the whole day in there and see nothing of London. If they've only got one day, I'd want them outside.”
“That's fair. The boat trip, then — you see half the city from the river in an hour.”
“Yes, and it ends at Greenwich, which is a nice place to walk anyway. I'm less sure about the stadium tour, honestly. It's great if you love football and boring if you don't.”
“True. And the market covers food and people at the same time, which the stadium doesn't.”
“So are we saying the boat and the market? The museum they can do next time.”
“Yes — one thing on the water, one thing on the ground. I think that's a good day.”
What usually went wrong in the one-take version. Two speeches instead of a conversation — each person talking for forty seconds without a question. Agreement with everything, so no reasons needed. And a decision that was never actually said: the recording stops with both people still listing. The four things being marked here, in the order they are most often lost: sharing the floor (ask twice), responding (That's fair… True… I hadn't thought of that), reasons (a because on every opinion), and reaching the decision out loud.
If you recorded it alone, the test is whether you changed direction at least twice — whether the second voice ever disagreed with the first. A conversation with no disagreement is a monologue with two names.
The examiner now asks you both some wider questions about the same subject — visiting places, travelling, seeing a country for the first time. You are giving opinions now, and answering each other as well as the examiner. Record all five in one take, about thirty seconds each. This is the part where the strongest candidates pull away, because it is the only part where you can bring in something from your own life that nobody could predict.
The debrief — read this after you have recorded. A model answer, and then what makes it one.
“Should you learn about a country's history before you visit? — Partly, I think. If you know nothing, a castle is just a big old building and you walk through it in ten minutes. But if you know too much, you spend the visit checking things instead of looking at them. When I went to Lisbon with my school, the person who enjoyed it most was my friend who had read one page about the earthquake — one page, not a book. So I'd say: learn one story, and let the place tell you the rest.”
Four moves, and you can hear each one: a position that is not simply yes or no (Partly); a reason on both sides with if; a concrete example from real life; and a conclusion that goes slightly further than the question asked. Any of the five questions above can be answered with those four moves.
What usually went wrong in the one-take version. One sentence per question. An opinion with no example — I prefer cities because they're more interesting, which tells the examiner nothing. And no reaction to the partner: Part 4 is a discussion, not two interviews in the same room. If your partner answers first, begin from what they said: I'd add something to that…, I see it slightly differently…, That's a good point, but…. Three phrases, and they lift this part more reliably than any vocabulary list.
The three sentences to have ready before you walk in. Sorry, could you repeat that? — when you did not hear. Do you mean…? — when you did not understand. That's an interesting question, actually. I suppose… — when you need three seconds to think, which is allowed, and which sounds far better than silence.
Four recordings, made in one take each. Listen to each one once for content — did you answer the question, and did you finish? — and once for sound: speed, pauses, the ends of words. Two listenings, two different jobs.
Under exam conditions the usual result is one part that ran out early — a Part 2 that stopped at thirty seconds, a Part 4 answer that was one sentence. That part is the one to record again tomorrow, three times, with the clock showing.