Three parts, sixteen minutes, taken in pairs, under exam conditions: start the clock, record each part in one take without stopping to think, and do not read the model until the recording exists. There is no Check button and no score, because a spoken answer is not something a web page can mark. The recording is the paper; the model 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 each candidate a few general questions. It is short, it is not a warm-up and it is assessed. Two or three sentences per answer: a direct response, a reason, and one concrete example. Under exam conditions: read the four questions once, start the clock, and answer all four in a single take of about two minutes. Then read the model.
In Part 1 the examiner is establishing a baseline — how you sound when you are not under pressure — and giving you an easy place to show natural language. The questions are ordinary. The answers should not be.
The commonest C2 mistake in this part is the prepared speech. An examiner can hear a rehearsed paragraph from the second sentence, and it costs marks under Interactive Communication because it is not a response to what was asked. One take, no notes, is the cure.
Examiner: Is it more important to be accurate or to be interesting?
Candidate: Accurate, and I say that as somebody who would much rather be the other thing. The reason is that interesting is cheap to produce and accuracy is not: anyone can be interesting for a minute by being slightly wrong in a memorable way, and most of what gets quoted is exactly that. Where I would qualify it is that accuracy nobody listens to is not worth much either — a true sentence that puts the room to sleep has failed at the one job a sentence has. So I suppose my honest answer is: accurate first, and then work on the sentence until it is interesting as well, rather than instead. The order matters. You can make a true thing interesting. It is much harder to make an interesting thing true.
Why this is a top-band answer, and it is only forty seconds long: a side in the first word, with an admission that makes it credible; a reason that is an argument rather than a preference; a qualification that does not collapse into “both”; and a closing pair of sentences whose grammar carries the point — the same structure twice, reversed. No introduction, no list, no “in conclusion”.
Debrief: play your own recording against this. Did your first word take a side? Did you give a cost? Did you stop, or trail off? Those three are the whole of Part 1.
The examiner gives you both a set of prompts and a question. You talk to each other for about two minutes, and then the examiner gives you a second question and you have about one minute to reach a decision. Under exam conditions: fifteen seconds to read the card, then one continuous take of four minutes. The model is for afterwards.
The situation: A company is designing a one-day induction for new staff arriving from abroad. Five sessions have been proposed for the day.
A · What your manager actually means. A session decoding common workplace phrases — “when you get a moment”, “I’d be inclined to…”, “not a priority”.
B · The email tone workshop. Rewriting real emails that were read as rude, and real emails that were read as insincere.
C · Lunch with a colleague. An unstructured hour with a member of staff who joined from abroad two years ago.
D · The compliment and the complaint. How to give and receive both without either being misread.
E · Reading the room. A guided walk round the office, noticing who sits where, who speaks first, and what silence means in a meeting.
First question (talk to each other, about 2 minutes): Here are some ideas for an induction day for staff arriving from abroad. Talk to each other about how useful each of these would be in helping newcomers avoid misunderstandings.
Second question (about 1 minute): Now decide which two sessions should be kept if the day is cut to a morning — and which one you would drop first.
A · Shall I start with the lunch, because I think it’s the one everyone will want to drop and I think it’s the one that does the most?
B · Go on — why the most?
A · Because every other session is somebody explaining the code. The lunch is an hour with somebody who cracked it two years ago and still remembers not having it. That’s the only person in the building who can tell a newcomer which of the phrases actually matter.
B · I’d push back on that a little. An hour with one person is an hour with one person’s prejudices. The manager-phrases session is at least systematic — “not a priority” means “no”, and you can learn that in a minute and use it the same afternoon.
A · That’s fair. But “not a priority” means “no” from one manager and “not yet” from another, and the session can’t tell you which. The walk round the office can — that’s the one I’d pair the lunch with, because they’re both about reading people rather than phrases.
B · Whereas the email workshop and the compliment session are both about producing language, not reading it — and a newcomer does far more reading than producing in the first month.
A · So if it’s cut to a morning: keep the walk and the lunch, drop the compliment session first, because it’s the one that’s really a subset of the manager session anyway?
B · I’d keep the manager session over the lunch, honestly — but I’ll concede the walk. Say: the walk and the manager phrases, and the compliment session goes first. Though we’ve just dropped the only one where anyone gets to practise.
What the examiners are hearing: an opening that takes a risk rather than the first item; a disagreement with a reason, answered with a distinction (“no” against “not yet”); the two candidates building a category together (reading against producing); a decision that is negotiated rather than announced — one side concedes the walk, the other concedes the lunch — and a closing line that shows the speaker is still thinking.
Debrief: in your own take, find the moment you disagreed. If there isn’t one, that is the note. Then find the moment you closed. If there isn’t one, that is the other note.
Each candidate gets a card with one question and three ideas, and speaks alone for two minutes. The partner then comments for about a minute, and the examiner asks both of you a further question. Then there is a five-minute discussion on the wider theme. This part carries the most marks in the paper. Under exam conditions: fifteen seconds with the card, then two minutes in one take, timed. The model is read afterwards, against your recording.
Candidate A’s card
What makes a piece of language memorable?
· its rhythm
· its element of surprise
· who says it
Candidate B’s card
Why do people find some jokes offensive and others harmless?
· the target
· the teller
· the setting
The examiner’s follow-up questions for the discussion — the examiner will choose from these, and will not warn you which:
· Is being memorable more important than being right?
· Can a joke be harmless and still be worth telling?
· Should there be any subject that cannot be joked about?
· Do we remember what was said, or who said it?
· Has the internet made language more memorable, or less?
A model two-minute long turn on Candidate A’s card: What makes a piece of language memorable?
I am going to argue that the honest answer is the least respectable one on the card — that it is mostly who says it — and then say why that is not quite as depressing as it sounds.
Take rhythm first, because it is the answer everyone gives and it is not wrong. The sentences that survive are the ones with a beat in them: the three items, the reversal, the stress on the last word. I can prove that rhythm does the work by pointing to the misquotations — the lines people remember are almost never what was said, they are what was said with the rhythm repaired. So rhythm is real. But rhythm on its own produces advertising, and nobody thinks advertising is memorable in the sense the question means; it is merely hard to get rid of.
Surprise I would treat more carefully, because I think it is the mechanism rather than the cause. A memorable line does arrive a half-beat before you expect it, or means something other than what it first seemed to — that is the pun, the punchline, the last line of the poem. But surprise wears off on the second hearing, and the lines that matter survive a thousandth hearing. Something else is holding them up.
Which is where I come to who says it, and I would use it not as a third item but as the thing the other two depend on. The same sentence with the same rhythm and the same surprise is forgettable from a stranger and permanent from a person who has earned the right to say it — a parent, a teacher, someone at a funeral. We do not remember sentences. We remember sentences from people, and the from is doing most of the work.
So: rhythm makes it repeatable, surprise makes it land, and the speaker makes it matter. If I had to keep one, I would keep the speaker, which is why the most memorable things anyone has ever said to me would look like nothing at all written down.
What makes it a top-band turn: a position in the first sentence, framed as a risk; the three prompts handled in a deliberate order, with the third promoted to the cause of the other two rather than listed alongside them; one piece of evidence (misquotation) offered as proof and not as decoration; a distinction between mechanism and cause; and a closing sentence that turns back on the speaker’s own life. The grammar does real work — a cleft, a fronted which is where, a conditional — and none of it was put there to be noticed.
Debrief: compare the timing of your position with this one’s. If yours arrived after thirty seconds, that is the first thing to fix; nothing later in the turn can recover the shape.
Nothing on this page is scored, so the numbers below stay empty. That is deliberate: Speaking is assessed by two people in a room, on Grammatical Resource, Lexical Resource, Discourse Management, Pronunciation and Interactive Communication.
The debrief question for an exam-conditions recording is not “was it good?” but “did it last?”. Play back Part 3 with a stopwatch. If the long turn ended before 1:45, or the discussion turns were under fifteen seconds, that is the whole diagnosis — and it is fixed by rehearsing endings, not by learning vocabulary.