IELTS speaking part 3: questions and the 45-second answer

The examiner takes the cue card back, and the test changes shape. For two minutes you described a shop you like, a teacher you remember, a trip that went wrong. Now the questions stop being about you. "Why do you think some people prefer small shops to big chains?" "Will physical shops still exist in twenty years?" That's part 3, and it's the stretch where interviews cruising at 6.5 either find the 7 or don't.

On paper it's the simplest part: a 4 to 5 minute discussion with no cue card, no prep time and no fixed script, built on the topic you just spoke about. In practice it's the part candidates prepare for least, because you can't memorise your way through it. This guide covers the register shift part 3 tests, a 45-second answer shape, the six question types and the grammar each one invites, and what to do when the examiner pushes back.

What part 3 is, and why it weighs more than its five minutes

The official format page describes part 3 as a discussion of issues "related to the topic in Part 2 in a more general and abstract way and, where appropriate, in greater depth," filling the last 4 to 5 minutes of an 11-to-14-minute test. ielts.org is unusually direct about what this section measures: "your ability to explain your opinions and to analyse, discuss and speculate about issues."

Read that against the other two parts. Part 1 asks for familiar facts. Part 2 asks for description. Analysis and speculation have exactly one place in the test to show up, and this is it.

There's no separate part 3 score. The examiner rates your whole interview on four criteria: fluency and coherence, vocabulary, grammar and pronunciation, then averages the four marks into one band. But the parts don't contribute equally to a borderline call. Band 7 grammar wants complex structures produced reliably, and the conditionals, comparisons and speculative forms that count as complex are exactly what part 3's questions invite. An interview sitting between 6.5 and 7 tends to be decided by the five minutes where that language either appeared or didn't.

The shift from "I" to "people"

Part 1 asked, "Do you like shopping?" Part 3 asks, "Why do you think some people enjoy shopping more than others?" Same territory. Different subject. The question is now about people in general, and the most common way to lose marks in part 3 is to keep answering about yourself.

Listen to the difference on "Why do so many people shop online now?":

Band 6: "For me, I prefer shopping online because it's cheaper and I don't have time to go to shops. Last month I bought my shoes online, and it was very easy."

Band 7: "Convenience, mostly. People work longer hours than they used to, so getting an hour back by ordering from your sofa is hard to beat. I'm the same — my last pair of shoes never saw a shop. Though I suspect price matters more than most of us admit."

The first answer is decent English wasted on the wrong question. It was asked about people and answered about me. The second still contains a personal detail, but as evidence for a general claim, not as the whole answer. That's the move: claim about people, example from your life. Run it in reverse, story with no claim, and the answer never engages the question at all.

The 45-second answer: position, reason, example, push

Part 1 runs on a 25-second shape: answer, reason, detail. Part 3 stretches it one move further: state a position, give a reason, ground it in an example, then push one step past it with a concession, a comparison or an implication. Three to five sentences, somewhere around 30 to 60 seconds. Then stop and let the examiner steer. Part 3 is a discussion, not a second long turn.

Here's the shape on a classic question, "Do you think advertising influences what people buy?":

Band 6: "Yes, I think so. We see many advertisements every day on TV and social media, so when we go shopping, we remember the products and we buy them."

Band 7: "To some extent, though less directly than it used to, I'd say. Hardly anyone watches an ad and rushes out to buy the product. It works more slowly than that, by making a brand feel familiar, so when I'm choosing between two coffees at a convenience store, I tend to grab the one I've seen before. That said, a friend's recommendation still beats any ad."

The band 7 answer opens with a hedged position ("to some extent"), explains a mechanism instead of asserting a fact ("making a brand feel familiar"), drops in one personal example as evidence, and closes on a concession. About 40 seconds of speech. Notice the graded certainty running through it: to some extent, hardly anyone, tend to. Hedging like this isn't weak English. The skill the test names is speculate, and would, might and tend to are the grammar speculation is made of.

Six question types, and the grammar each one invites

You can't predict part 3 topics, but the question types repeat, and each one is an invitation to a specific structure. The same way part 1 questions target tenses, part 3 questions target complexity:

• Opinion. "Do you think advertising is a bad thing?" Invites a stance with reasons. The trap is a bare yes; position language ("on balance," "to some extent") is what registers.
• Comparison with the past. "How has shopping in your country changed?" Invites used to, the present perfect, and comparatives: "far fewer cash payments than a decade ago."
• Prediction. "Will small shops still exist in twenty years?" Invites will, be likely to, might, and graded certainty between them.
• Hypothetical. "If advertising aimed at children were banned, what would change?" Invites the second conditional: would and might plus a base verb.
• Cause and effect. "Why do you think people buy things they don't need?" Invites causal chains: "one reason is," "partly because," "which leads to."
• Problem-solution. "What could governments do about overconsumption?" Invites should and could with impersonal subjects: "one option would be to..."

You don't need to identify the type in real time. Drill the structures, and the right one is available when its invitation arrives. If conditionals or comparatives collapse under pressure, that's a grammatical range problem with a name, and it's fixable in isolation.

When the examiner pushes back

In part 1 the examiner works through a script and can only repeat a question, never rephrase it. Part 3 is the opposite: the examiner is free to follow your answer with "Why?", "Is that always the case?", or "What about older people?" The format is built to go deeper where your answers allow it. Probing means the discussion is working, not that your last answer failed.

Four rules for the pushback:

A "why?" is a request for a second reason. Repeating your first reason with more confidence adds nothing scoreable; a new angle does.

Disagreeing is allowed, including with the question's premise. "Do you think technology makes people lazy?" can be answered with "I'm not sure it does, to be honest..." and a defended no outscores a hollow yes, because the test marks language, not opinions.

You can ask for a rephrase. If the question didn't land, "Do you mean whether this is a problem for society?" costs nothing, and clarification language is itself speech the examiner can credit.

No knowledge is not no answer. "I honestly don't know much about farming, but I'd guess..." is a fully valid opening, because speculating past the edge of what you know is the named skill of this section. Invented statistics are not: "In my country, 80% of people..." sounds informed for exactly as long as the follow-up question takes to arrive.

What the four criteria hear in part 3

Fluency and coherence. Hard questions produce the longest silences of the test. A thinking-aloud phrase, "That's not something I've thought about, but...", is real language that holds the floor; dead air isn't. Use it once or twice, not before every answer, or it turns into the memorised filler examiners discount.

Lexical resource. Abstract questions reward topic-neutral abstract words: trade-off, incentive, trend, side effect. The cheaper habit is paraphrase: if the question says "influences what people buy," an answer that says "shapes purchasing decisions" has just produced lexical resource on record. Echoing the question's own words back produces nothing.

Grammatical range and accuracy. This is the criterion part 3 was built for. Conditionals, speculative modals and comparative structures are the "complex sentence forms" the band 7 descriptor is looking for, produced live under pressure.

Pronunciation. Cognitive load flattens delivery. While you're constructing an opinion, the stress and intonation that carried your part 2 story tend to disappear, so keep placing weight on the meaning-carrying words ("it works more slowly than that").

For the band 6 versus band 7 line on each axis, the speaking band descriptors guide breaks all four down.

Mistakes that cost 0.5 in part 3

• The anecdote-only answer. A story about your cousin is evidence without a claim. Open general, then illustrate.
• The reflex agreement. "Yes, I agree, advertising influences people a lot" restates the question and adds zero new language.
• Essay register out loud. "Moreover, it is undeniable that..." belongs in Writing Task 2. Part 3 runs on spoken precision: "that said," "the way I see it."
• Part 1-sized answers. "Maybe in the future, yes." One clause can't hold a position, a reason and an example.
• The 90-second monologue. Part 3 is a dialogue. Land the point and hand it back; the examiner cutting in usually means they have what they need on that question.
• Inventing facts to sound informed. A made-up statistic impresses nobody and invites a follow-up you can't answer. Hedge honestly instead; hedging is the assessed skill.

Drill part 3 in five minutes a day

A full mock interview gives you one part 3 per half hour, with your attention split three ways. Five-minute drills isolate the exact muscles this section uses:

• The zoom-out rep. Take any part 1 question and re-ask it at society level: "Do you like shopping?" becomes "Why do people in your country shop online so much?" Answer in 45 seconds with the shape: position, reason, example, push.
• The hedge ladder. Take one prediction question and answer it three times at three levels of certainty: will, is likely to, might. You're drilling speculation grammar on purpose instead of hoping it appears.
• The pushback rep. Record a 45-second answer, then immediately ask yourself "Is that always true?" and answer with a second reason. The follow-up muscle is trainable; most people have just never trained it.

One rep, one focus, every day. That's the IELTS Drills approach applied to the hardest five minutes of the interview: each drill's feedback is tagged to a single Band Descriptor, so if collapsing conditionals are what's quietly capping your grammar score, you find out, and you drill that instead of "speaking" in general. Not sure which of the four axes is your lowest? Start by naming it.

Frequently asked questions

How long should a part 3 answer be? Around 30 to 60 seconds, three to five sentences. Long enough to hold a position, a reason and an example; short enough to stay a conversation. The two-minute material belongs in part 2.

How many questions are there in part 3? No fixed number. Part 3 runs on time, not question count: 4 to 5 minutes, which usually means somewhere between four and eight questions depending on how developed your answers are.

Can I ask the examiner to rephrase a question? Yes. Unlike part 1, where questions can only be repeated, part 3 allows rephrasing, and asking ("Do you mean...?") is itself scoreable clarification language. Once or twice costs nothing.

What if I know nothing about the topic? Say so, then speculate: "I don't know much about urban planning, but I'd guess..." The section explicitly tests your ability to speculate, so reasoning past your knowledge is the skill, not a confession of failure.

Is it OK to disagree with the examiner's question? Completely. The examiner has no opinion to offend; the questions are prompts. A defended disagreement produces more scoreable language than polite agreement.

Does part 3 have its own band score? No. All three parts are scored as one performance on the four criteria. Part 3 just tends to supply the evidence that settles borderline calls, because it's where complex language gets invited.

The bottom line

Part 3 is the only five minutes of the test that asks you to think about the world out loud, and it rewards a specific set of moves: a general claim before the personal story, a position-reason-example-push shape that lands inside 45 seconds, hedged speculation instead of invented certainty, and a second reason ready for the "why?".

None of that is a personality trait. It's six question types and a handful of structures, and five minutes a day of zoom-out reps and hedge ladders covers the lot inside two weeks. The interview that ends with a real discussion instead of a polite Q&A is the one that comes back 0.5 higher.