Part 3 is the first point in the Listening test where the answer is not a fact — it is a person. Up to now you have been catching phone numbers and following one voice around a map. Now the test puts multiple voices in a room and lets them argue. The question is rarely "what was said" — almost everything gets said — it is "who was left holding which view when the conversation moved on." That is a different skill from hearing, and it is the one this part is built to test.
What Part 3 officially is
Here is ielts.org's description:
"Parts 3 and 4 deal with educational and training situations. In Part 3 there is a conversation between two main speakers (for example, two university students in discussion, perhaps guided by a tutor)."
So the setting is academic — a seminar, a project meeting, a tutorial — and the mode is discussion. Like every Listening recording, it plays once:
"You will hear the recordings once only."
Part 3 carries 10 of the test's 40 questions — ielts.org states the Listening section "has four parts, with 10 questions in each part" and "Each correct answer receives 1 mark." There is no extra weighting for the harder parts: a missed attribution here costs exactly what a missed phone number cost in Part 1.
How many speakers? Officialdom disagrees
You would expect the two IELTS partners to agree on how many people are talking. They do not. ielts.org says:
"In Part 3 there is a conversation between two main speakers (for example, two university students in discussion, perhaps guided by a tutor)."
The British Council's format page describes the same recording as:
"a conversation between up to four people set in an educational or training context"
Two main speakers, or up to four people — the official sources give different numbers, and we are not going to pretend they agree. The safest reading is that the conversation centres on two main speakers but can include more voices — a tutor guiding two students is the classic shape. Prepare for the possibility of a third or fourth voice rather than being surprised by one.
The failure: attribution
Call it the attribution trap. In Part 3, wrong options are almost never inventions — they are things genuinely said in the recording, attached to the wrong person, or abandoned by the person who said them. Three versions of the trap recur:
- Mentioned but rejected — a speaker raises an idea precisely so it can be knocked down.
- The mind-changer — a speaker starts with one position and ends with another; the question scores their final position, not their opening one.
- Wrong owner — the view is real, the words are real, but they came out of a different mouth.
If you answer on "I heard that", you will hit all three. You have to track ownership, in real time, once only.
A worked example
Coaching material: the scenario below is our own illustration, written for this page. It is not an official IELTS recording or an official transcript.
Tutor Dr Okafor is meeting two students, Lena and Marcus, about their field project on city allotment gardens.
Dr Okafor: So — how will you collect the data? Questionnaires? Marcus: We considered questionnaires, but honestly the response rates would be terrible. Gardeners are outdoors — they won't sit filling in forms. I'd rather do short interviews on site. Lena: Interviews worry me. They take ages, and we've only got three weekends. My instinct was to just observe and count — plot sizes, what's growing, who turns up. Dr Okafor: Observation alone won't tell you why people garden, which is your research question. I'd suggest combining methods. Lena: Hmm. Actually, if we split up — Marcus interviews while I observe — the time problem disappears. Fine, I'm persuaded. Interviews plus observation. Marcus: Agreed. Though I still think we should keep the interviews under ten minutes.
Q1. What does Marcus think about questionnaires? A) They would produce too few responses. B) They are the quickest method. C) They suit outdoor research.
The trap is that questionnaires are discussed warmly enough for B and C to feel plausible — the tutor is the one who floats them. But Marcus's words are "the response rates would be terrible": questionnaires are mentioned only to be rejected, and A is the answer. If you tick an option because its topic came up, you have fallen for exactly what this part is testing.
Q2. What data collection method does Lena support by the end of the discussion? A) Observation only. B) Interviews combined with observation. C) Questionnaires.
A is Lena's own opening position — "just observe and count" — spoken in her own voice, so it is the most dangerous option on the page. But she is the mind-changer: after the tutor's objection she says "Fine, I'm persuaded." The question asks where she ended, not where she started — B. Part 3 loves the phrase "by the end" precisely because openings are quotable and endings win.
Q3. Who suggests combining two methods? A) Marcus. B) Lena. C) Dr Okafor.
Lena is the one who says "interviews plus observation" out loud and works out the logistics, so B feels right — she owns the plan by the end. But the suggestion itself is the tutor's: "I'd suggest combining methods." The view is real; the trick is the owner — C. Adopting an idea is not the same as suggesting it, and Part 3 questions are written to punish anyone who blurs that line.
What is not official here
The three trap patterns above — rejected options, mind-changers, wrong owners — are our coaching labels, drawn from how discussion recordings behave, not terms you will find on ielts.org. The official pages verify the setting, the once-only playback, the 10 questions and 1 mark each, and the six Listening question types by name ("Multiple choice", "Matching" and four others) — but publish no official description of what makes Part 3 questions hard, and, as quoted above, do not even agree on the speaker count. Treat any page that states "Part 3 always has exactly N speakers" with suspicion — including this one, which is why we quoted both sources instead.
What actually costs you
- Answering on recognition. Every option was probably said. "I heard it" proves nothing.
- Locking in a speaker's first position. The recording plays once — if someone concedes late, your early answer is wrong and there is no second pass.
- Losing track of voices when a third one joins. British Council says up to four people; if you have only ever practised two-voice dialogues, a tutor's interjections will scramble your attribution.
- Spelling and word limits still bite. ielts.org: "you will lose marks for incorrect spelling and grammar", and word limits such as "NO MORE THAN TWO WORDS AND/OR A NUMBER" apply here as everywhere.
Try one
[Try a Listening task →] — one full task, marked, no card required.