The Real World Problem
Every exam period produces the same confused candidate. They speak English every day at work. Their colleagues understand them without effort. They walked out of the Speaking room feeling comfortable. And their Speaking score came back lower than papers they were far more worried about.
The explanation is almost never their English. It is that they treated the Speaking test as one performance, when it is actually four separate tasks, each designed to elicit different language and each rewarding different behaviour. Candidates who prepare "speaking" in general terms leave marks on the table in every part. Candidates who know what each part exists to measure prepare differently, and score differently.
This article completes our series on the Cambridge papers. We have already covered how Listening engineers its distractors and what Writing examiners actually reward. Speaking follows the same logic as Writing: fixed criteria, applied consistently, mostly invisible to candidates who have never read them.
Before the four parts: who is marking you, and against what
Two things frame everything that follows.
First, there are two examiners in the room, and they are not doing the same job. The interlocutor speaks to you, manages the tasks, and awards a single Global Achievement mark. The assessor sits apart, says nothing, and does the detailed work: marking you against the analytical criteria. At B2 First these are Grammar and Vocabulary, Discourse Management, Pronunciation, and Interactive Communication. At C1 Advanced, grammar and vocabulary become separate scales, so the bar for lexical range is explicitly higher. The practical consequence is that the person you instinctively perform for is not the person marking you in detail, and in the interactive parts of the test, playing to the interlocutor instead of your partner works directly against the criteria.
Second, three of those criteria translate into behaviour worth naming before we walk through the parts. Grammar and Vocabulary rewards range over safety: a candidate who attempts conditionals, passives, and comparatives with occasional errors demonstrates more than one who is flawless in the present simple. Pronunciation assesses intelligibility, not accent: a Spanish or Catalan accent costs nothing, while flat intonation or misplaced word stress that obscures meaning does. And Interactive Communication measures how you build conversation with another person, which is the criterion most candidates never practise, because most preparation happens alone or with a teacher who accommodates them.
With that frame in place, here is the test, part by part. The official format runs to roughly 14 minutes per pair at B2 First and 15 at C1 Advanced.
Part 1: the interview, where marks are lost through minimalism
Part 1 is around two minutes of direct questions about you: your work, your studies, your plans, your interests. It is deliberately the easiest part of the test, and that is exactly the trap.
Because it feels like small talk, candidates treat it as a warm-up and give minimal answers. The examiners are already marking. "Do you enjoy your job?" answered with "Yes, it is interesting" scores what it deserves: it is grammatically accurate, communicatively empty, and gives the assessor nothing to place on the Discourse Management scale.
The fix is mechanical and costs nothing. Extend every answer by one reason or one example. "Yes, I enjoy it because no two projects are the same, and I like solving problems under pressure." One habit, applied consistently, turns Part 1 from a wasted two minutes into free evidence of extended, coherent speech before the harder tasks begin.
Where marks are lost in Part 1: one-line answers, memorised self-introductions that ignore the actual question, and answers so rehearsed they collapse the moment the question deviates from the script. Rehearsed chunks are easy to detect and fail the relevance test instantly.
Part 2: the long turn, where marks are lost by describing instead of comparing
Part 2 gives you two photographs (three at C1 Advanced, of which you discuss two) and one minute to speak alone, followed by a short response to your partner's photographs. Two failures dominate this part, and between them they account for most of the marks lost here.
The first is describing instead of comparing. The task instruction asks you to compare the photographs, but anxious candidates default to inventory mode: "In the first photo I can see a man, he is in an office, there is a computer." Description fills the minute without demonstrating the language the part exists to elicit. Comparison and speculation language is the entire point: "whereas," "while," "both images show," "the people in the second photo might be," "it looks as though."
The second is ignoring the printed question above the photographs. There is always one, and it is the actual task. A candidate can compare fluently for a full minute and still lose relevance marks because they never addressed the question they were asked to answer.
The third, smaller leak: the 30-second response to your partner's photographs. Candidates switch off once their own minute ends. That short response is still assessed, and a thoughtful sentence there is one of the cheapest marks in the test.
One minute is shorter than candidates expect and longer than their prepared material. Recording yourself doing timed long turns tells you within two attempts whether you compare or merely describe.
Part 3: the collaborative task, where marks are lost by dominating or racing
Part 3 gives you and your partner written prompts arranged around a central question, with a discussion phase and then a decision phase. This is the Interactive Communication showcase, and it is where the most persistent myth in Cambridge preparation does its damage. That myth gets its own section below, because it deserves one. First, the two rules that govern the task itself.
Rule one: do not try to cover every prompt. The prompts are material, not a checklist. Depth on three beats a superficial tour of five, because developed discussion generates the extended, connected speech that Discourse Management rewards, while racing through prompts generates fragments.
Rule two: do not agree instantly. The task is engineered to produce negotiation. Instant agreement ("Yes, I think so too. Next one?") kills the interaction the part exists to assess. Polite disagreement ("I see your point, but I am not sure, because...") generates exactly the language the examiners are listening for. Reaching the final decision matters far less than how you negotiate towards it. Two candidates who disagree constructively and run out of time before deciding have outperformed two who agreed on everything in ninety seconds.
Where marks are lost in Part 3: monologuing at the interlocutor instead of talking to your partner, treating the prompts as a list to complete, agreeing with everything, and the domination problem covered next.
The partner myth
Now the question every candidate asks, usually in the final week: what if I get a bad partner?
Here is the mechanism, because understanding it dissolves the anxiety. You are assessed individually, against fixed criteria, not against or relative to the person beside you. A partner with weaker English does not drag your score down. A partner with stronger English does not make you look worse. The assessor is not comparing you; they are placing each of you independently on the same bands.
What a difficult partner creates is not a scoring problem but an interaction problem, and how you handle it is itself assessable behaviour that can work in your favour. If your partner gives short answers, inviting them in ("What do you think about this one?") is precisely the initiating and turn-management behaviour that Interactive Communication rewards. If your partner dominates, finding a natural entry point demonstrates the same skill from the other direction.
The behaviour that genuinely damages your score is the one anxious candidates adopt deliberately: dominating the collaborative task to "show your level." Talking over your partner, ignoring their contributions, and steering every exchange back to prepared material reads to the assessor as an inability to interact, which is exactly the criterion Part 3 exists to measure. Candidates spend hours worrying about the partner they cannot control and no time practising the interactive behaviour they can.
Part 4: the discussion, where marks are lost through unsupported opinions
Part 4 extends the Part 3 topic into broader, more abstract questions, around four minutes at B2 First and five at C1 Advanced. This is where the level ceiling is tested: opinion, justification, speculation, abstraction. Questions may be directed at one candidate or opened to both.
Two leaks dominate. The first is the unsupported opinion: "Yes, I think technology is good for education" and then silence. Part 4 questions are invitations to justify, and the justification is where the marks live. A useful internal rule: no opinion without a "because," and where possible, a "for example."
The second leak is failing to use your partner. When a question is opened to both candidates, building on the previous answer ("I agree with what she said about cost, and I would also add...") is high-value interactive behaviour that almost nobody deploys, because almost nobody prepares Part 4 as a conversation. It is usually rehearsed, if at all, as a solo opinion drill.
Where marks are lost in Part 4: one-sentence opinions, retreating to safe simple structures precisely when the test is inviting complex ones, and treating the final minutes as an individual interview when they are the last chance to demonstrate interaction.
What this means for your preparation
If Speaking is one of your target papers, three adjustments follow directly from the structure above.
First, practise with a partner, not only a teacher. A teacher accommodates you; a peer does not. Interactive Communication cannot be rehearsed in a format where the other person manages the conversation for you.
Second, record timed Part 2 long turns. The recording answers, immediately and without flattery, the only question that matters in that part: are you comparing, or describing?
Third, stop rehearsing speeches and start rehearsing functions: comparing, speculating, disagreeing politely, inviting an opinion, buying thinking time. Functions transfer to any prompt. A memorised paragraph transfers to none.
And keep the paper in proportion. Your final grade is one overall Cambridge English Scale score built from five equally weighted components, with the B2 boundary at 160 and C1 at 180, and no minimum required in any individual paper. Speaking is one fifth of the outcome. Whether it is the fifth you should spend your remaining weeks on is not something an article can tell you, because it depends on your profile, not on the exam. That is he difference between knowing your level and knowing your gaps , and it is what a proper diagnostic exists to establish. And if you are still deciding which exam to sit at all, start with B2 First versus C1 Advanced.
Frequently asked questions
Does my partner's level affect my Cambridge Speaking score?
No. Examiners assess each candidate individually against fixed criteria, not against each other. A stronger or weaker partner does not raise or lower your score. What can cost you marks is how you handle the interaction: dominating the conversation or failing to respond to your partner both damage your Interactive Communication score.
Does my accent matter in the Cambridge Speaking test?
No. The Pronunciation criterion assesses intelligibility: whether your stress, intonation, and individual sounds allow the examiner to understand you easily. A Spanish, Catalan, or any other accent is not penalised. Marks are lost when pronunciation interferes with understanding, not when it reveals where you are from.
Can I fail the Speaking paper and still pass B2 First?
Yes, in principle. Your result is one overall Cambridge English Scale score built from five equally weighted components, and there is no minimum required in any single paper. A weak Speaking score can be compensated by stronger scores elsewhere, provided your overall score reaches the B2 boundary of 160. Relying on that compensation is a risky strategy, but a bad Speaking day does not automatically fail you.
How long is the B2 First Speaking test?
Around 14 minutes per pair of candidates, across four parts: a short interview, an individual long turn with photographs, a collaborative task with written prompts, and a final discussion. The C1 Advanced version follows the same four-part structure and lasts around 15 minutes.
What happens if I do not understand the examiner's question?
Ask for repetition. A natural request such as "Sorry, could you repeat that?" does not lose you marks; it is normal communicative behaviour and examiners expect it. What costs marks is silence, or answering a question you invented because you did not want to ask. Asking once is fine. Needing every question repeated will start to affect the impression of your listening comprehension.
Can I take the Cambridge Speaking test alone?
Not by choice. The test is designed as a paired format because interaction is one of the assessed criteria. If a session has an odd number of candidates, the final group is arranged as a group of three, with timings adjusted. You cannot request an individual test.