The Cambridge Examiner Who Talks to You Marks One Thing

Most candidates walk into a Cambridge Speaking test and treat one examiner as the exam and the other as furniture.

Cambridge publishes free teacher guides explaining how the marking works, and they say the opposite. The examiner talking to you gives a single mark. The silent one gives all the rest.

SpeakShark is free to practise what the silent examiner is actually listening for, three AI conversation sessions a day with no card.

In this guide: the split · the numbers · what it changes · no chapter · a slip in the C2 guide · groups of three · limits · method · FAQ

Key takeaways

• Cambridge states there are two speaking examiners, with different roles, at every level we checked.
• The interlocutor speaks to you and marks Global Achievement only.
• The assessor stays quiet and marks everything else, four criteria at B1 and B2, five at C2.
• None of the three teacher guides has a chapter on Global Achievement, the interlocutor's mark.
• The C2 guide's glossary names a criterion that C2 does not use.

Two examiners, two completely different jobs

The wording appears in Cambridge's key terminology table, and it is the same sentence in the B1 Preliminary guide, the B2 First guide and the C2 Proficiency guide:

"The assessor listens and gives marks for Grammar and Vocabulary, Discourse Management, Pronunciation, and Interactive Communication."

"There are two speaking examiners in the test. The interlocutor speaks to the candidates and gives a mark for Global Achievement."

Read those two lines next to each other and the room reorganises itself.

The interlocutor is the examiner you meet. They greet you, ask your name, run the parts, hand you the photographs, tell you when to stop. Every social signal you receive during the test comes from this person. They mark one criterion.

The assessor sits to the side. They do not talk to you. They do not react. They mark four criteria at B1 Preliminary and B2 First, and five at C2 Proficiency where the grammar and vocabulary criterion splits into Grammatical Resource and Lexical Resource.

One mark against four or five

Put it in a table, because the proportions are the point.

One against four at the lower two levels. One against five at C2.

So the person whose approval you can read in real time is responsible for roughly twenty per cent of your speaking profile, and the person giving you nothing to read is responsible for the other eighty.

That matches what the published score tables show. When we went through 23 candidates' real examiner scores, Global Achievement was one column of five or six, and it was almost never the column that separated candidates from each other.

What the split changes in the room

We are not going to invent a rule about where to look. Cambridge does not publish one, and a guide that tells you to stare at the silent examiner would be making things up.

What the split does justify is a change in what you optimise for.

• Rapport is not the exam. Being liked by the interlocutor moves one mark. It is a real mark and worth having, but it is not the test.
• Silence from the assessor is not a bad sign. They are not supposed to react. Candidates read the lack of feedback as disapproval and tighten up, which costs them on the criteria that person is actually marking.
• Your partner is part of your score. Interactive Communication is an assessor criterion, and it is measured on what you do with the other candidate, not with the examiner. Cambridge's own commentary rewards candidates who ask their partner questions, as we found in the B1 case of Kenza and Mohammed.
• Nothing you say is off the record. Everything the interlocutor hears, the assessor hears too, and marks against four or five criteria rather than one.

The parts where this bites hardest are the collaborative ones, because that is where you are talking to a peer rather than to the examiner running the test. B2 First Part 3, C1 Advanced Part 3 and the C2 collaborative task all put you in that position, and B1 Preliminary Part 3 does it at the lowest level we checked.

Cambridge writes no chapter on the talking examiner's mark

Here is the detail that convinced us the split is deliberate rather than incidental.

Each guide has a contents page listing a chapter per assessment criterion. Here is what they cover:

• B1 Preliminary: Grammar and Vocabulary, Discourse Management, Pronunciation, Interactive Communication.
• B2 First: Grammar and Vocabulary, Discourse Management, Pronunciation, Interactive Communication.
• C2 Proficiency: Grammatical Resource, Lexical Resource, Discourse Management, Pronunciation, Interactive Communication.

Every chapter in all three guides covers a criterion the assessor marks. Not one of the three has a chapter on Global Achievement.

Cambridge wrote a teaching guide for every mark given by the examiner who says nothing, and none for the mark given by the examiner who talks. Whatever the reason, a teacher preparing students from these documents is being pointed entirely at the silent examiner's criteria.

The C2 guide names a criterion C2 does not have

While comparing the three guides we found something that looks like a copy and paste that was never updated, and it is worth flagging because it could mislead a teacher.

The C2 Proficiency guide's key terminology table says the assessor gives marks for "Grammar and Vocabulary, Discourse Management, Pronunciation, and Interactive Communication".

C2 Proficiency does not have a Grammar and Vocabulary criterion. Its own contents page, four pages earlier in the same document, lists Grammatical Resource at page 20 and Lexical Resource at page 26 as separate chapters. The split is the whole structural difference between the lower levels and C1 and C2, and we confirmed it against Cambridge's published examiner score tables, where C1 and C2 candidates receive six marks and B1 and B2 candidates receive five.

So the glossary in the C2 guide describes the B2 criteria. It names one subscale C2 does not use and omits two that it does.

This is a glossary slip rather than an error in the assessment scales, and it does not change how anyone is marked. It is worth knowing about because a teacher reading that table would take away the wrong list of criteria for the exam the document is about.

The subscale counts add to the confusion. The B1 guide says the scale is "divided into the four subscales", the B2 guide says five, and the C2 guide says five, while the published score tables show five marks at B1, five at B2 and six at C2. We could not read the reference table each guide points to, so we are reporting the wording rather than reconciling it.

A group of three gets a longer test, not a thinner one

One more useful thing sits in the same guides. Speaking tests can run with a pair or, when numbers are odd, a group of three, and candidates often assume three people means less airtime each.

Cambridge lengthens the test instead. The B2 First guide gives 20 minutes for a group of three across four parts. The C2 Proficiency guide gives 24 minutes for a group of three across three parts, and states directly that "the timing is adjusted for groups of three so that candidates have the same opportunities to speak".

So being put in a three is not a disadvantage in time. It does change the interaction, because there are two other people to respond to rather than one, and Interactive Communication is an assessor criterion.

If you want that skill sharper before test day, it is a conversation problem rather than a study problem. Role play scenarios and a daily speaking partner build the habit of responding to something you did not plan for, and testing your speaking level free tells you where you are starting. CEFR levels for speaking explains what the bands behind these criteria describe.

What we could not verify

We did not read the full assessment scales. These guides quote band descriptors and we are deliberately not reproducing them. Our claims are about who marks what and what the guides contain.

We could not read the pair timings. The published totals for pairs were lost in text extraction where a digit sits against a ligature, so we quote only the group of three figures, which extracted cleanly.

We could not resolve the subscale count wording. Four, five and five across the three guides against five, five and six actual marks. Each guide points to a table on its page 9 that we could not extract.

We did not check C1 Advanced. Cambridge publishes a guide for it and we read B1 Preliminary, B2 First and C2 Proficiency. The C1 criteria match C2's structure in the score tables, but we did not confirm C1's guide wording.

We are not claiming the assessor is more important. Global Achievement is a real mark from a trained examiner. The point is the proportion, not that one examiner matters and the other does not.

We did not verify current editions. The B2 guide's internal file name carries a May 2025 marker. Cambridge may have revised any of the three since.

How we researched this guide

We downloaded all three teacher guides from Cambridge's own preparation pages and extracted their text. Each file is around twelve megabytes, and the useful prose sits inside a great deal of embedded font data, so the extraction had to be filtered down to readable runs before anything could be read.

The finding came from reading the same section of three separate documents against each other rather than from reading any one of them. The assessor and interlocutor sentences sit in a key terminology glossary near the front, framed as vocabulary help for teachers rather than as a statement about how marks are distributed. In one document it reads as a definition. In three, it reads as a rule.

We then checked the claim rather than trusting one extraction. The assessor sentence and the interlocutor sentence were matched independently in each of the three files, and both returned the identical wording in all three, with the single difference that C2 writes "Global achievement" in lower case.

The C2 glossary slip surfaced the same way. We were comparing criterion lists across levels to build the table above, and the C2 terminology list did not match the C2 contents page in the same file.

Cambridge's site serves these PDFs normally. We fetched them from within a browser session on the same origin, which is our default for this site after several command line fetches returned nothing useful on pages that render perfectly well in a tab.

Practise speaking, from SpeakShark

SpeakShark is an AI English speaking practice app, and this finding is a fair argument for practising with something that does not react warmly to you.

Four or five of your marks come from an examiner giving you no encouragement at all. Practice that always comes with a friendly nod trains you for the one mark, not the four.

You talk, the AI answers what you actually said, and you get speaking feedback while the conversation is still running. The free tier gives basic feedback; the detailed pronunciation and grammar breakdown is on Premium.

Being straight about the limits: a free session runs five minutes with four turns, and it is one on one. It does not reproduce a room with a partner, an interlocutor and a silent assessor. For the collaborative parts you need people.

Start free with three sessions a day and no card. Paid sessions run ten minutes with unlimited turns. Limits are on the pricing page, and how it works walks through a session. Sign up here if you want the four marks rather than the one.

We are a speaking improvement tool. We are not an exam preparation provider, and we are not affiliated with Cambridge English, the IELTS partners, Pearson or any exam board. For test format, booking and official practice material, go to the exam body directly. Use SpeakShark to make your spoken English stronger, and use official material to learn the test.

• Site: speakshark.com
• Support: speaksharksupport@gmail.com
• Free tier: 3 AI conversation sessions per day, no card

Sources

• Cambridge English, B1 Preliminary and B1 Preliminary for Schools, Assessing speaking: a guide for teachers: the assessor and interlocutor definitions, the four subscales wording, the contents list of criterion chapters and the four part structure
• Cambridge English, B2 First and B2 First for Schools, Assessing speaking: a guide for teachers: the same two definitions, the five subscales wording, the contents list and the 20 minutes for a group of three figure
• Cambridge English, C2 Proficiency, Assessing speaking: a guide for teachers: the same two definitions including the Grammar and Vocabulary wording in its glossary, the Grammatical Resource and Lexical Resource chapters in its contents, the three part structure and the 24 minutes for a group of three figure with the equal opportunities statement
• Cambridge English, B1 Preliminary preparation: the page publishing the B1 guide alongside the examiner commentaries
• Council of Europe, CEFR Table 1, common reference levels global scale: the level scale these qualifications report against

FAQ