IELTS Speaking is an 11–14 minute interview with a certified examiner, marked on four criteria of equal weight. Two of them — Fluency and Pronunciation — do not exist in Writing at all, and they are where the familiar advice from essay practice stops applying. The most common pattern in a band 6.0 Speaking result is strong Lexical Resource and Grammar dragged down by hesitant delivery, not weak English.
What the examiner is doing in each part
All four criteria are marked across the whole interview, but each part is designed to create the evidence for a different one:
| Part | Format | What it exposes |
|---|---|---|
| Part 1 (4–5 min) | Familiar questions about you, your home, work or studies | Whether fluent everyday speech is automatic |
| Part 2 (3–4 min) | Cue card, 1 minute to prepare, speak for 1–2 minutes | Whether you can sustain and organise a long turn alone |
| Part 3 (4–5 min) | Abstract discussion linked to the Part 2 topic | Range of grammar and vocabulary under real cognitive load |
This is why Part 3 is where most candidates' band is actually decided. Part 1 answers can be rehearsed; Part 3 asks you to speculate, compare and justify in real time, and that is where complex structures either appear naturally or do not appear at all.
Fluency and Coherence
Fluency is not speed. A fast talker who repeats, restarts and self-corrects constantly scores below a measured speaker who keeps moving. What the descriptors penalise is hesitation to find language — searching for the word or repairing the grammar — as distinct from pausing to think about the idea, which is normal and unpenalised.
What raises it:
- Extending every answer. Even in Part 1, answer plus one reason plus one detail. One-line answers give the examiner nothing to mark and force them to work harder for evidence.
- Filling the two minutes in Part 2. The cue card bullets are a ready-made structure; use them as your paragraphs.
- Discourse markers used naturally — the thing is, having said that, what I mean is — which buy thinking time and signal structure at the same time.
- Moving on from a mistake. Repairing every slip fragments delivery more than the original error did.
| Band 6 sounds like | Band 7 sounds like |
|---|---|
| Willing to speak at length, though repetition, self-correction or hesitation is noticeable. Uses a range of connectives, but not always appropriately. | Speaks at length without noticeable effort. Some hesitation is language-related but occasional, and topic development is coherent with flexible use of discourse markers. |
Lexical Resource
Speaking rewards something Writing does not: idiomatic language. Natural collocations and common idiomatic expressions used accurately are explicitly credited from band 7 upward — provided they fit. Forced idiom is worse than none, because it reads as memorised.
What raises it:
- Paraphrasing when a word escapes you. Talking around a gap smoothly is itself evidence of resource; freezing is not.
- Precise everyday vocabulary. Cramped, run-down, overpriced rather than not very good.
- Topic vocabulary that arrives naturally in Part 3, where the questions turn abstract.
- Avoiding memorised set-pieces. Examiners are trained to detect rehearsed chunks, which are discounted.
| Band 6 sounds like | Band 7 sounds like |
|---|---|
| Has enough vocabulary to discuss topics at length and make meaning clear despite inappropriacies. Generally paraphrases successfully. | Uses vocabulary flexibly to discuss a variety of topics; uses some less common and idiomatic items with awareness of style, though with occasional inaccuracy. |
Grammatical Range and Accuracy
The same range-versus-accuracy trade-off as Writing, but under far more pressure — you cannot revise speech. What separates band 7 is that complex structures appear and most sentences come out clean.
What raises it:
- Answering hypothetical questions hypothetically. Part 3 questions such as "what would happen if…" are an explicit invitation to use conditionals. Answering them in the present simple leaves range unproven.
- Tense variety. Past for anecdotes, present perfect for experience, future forms for predictions — often all three in one Part 2 answer.
- Subordination: although, whereas, which means that. These are what turn two simple sentences into one complex one.
- Accuracy in the basics — third person -s, articles, plurals — which slip most when speed rises.
| Band 6 sounds like | Band 7 sounds like |
|---|---|
| Uses a mix of simple and complex structures with limited flexibility. Errors are frequent in complex structures, though they rarely block meaning. | Uses a range of complex structures with some flexibility. Frequently produces error-free sentences, though some grammatical mistakes persist. |
Pronunciation
The most misunderstood criterion. It does not measure how close you sound to a native speaker, and accent is not penalised — the descriptors are about intelligibility and the range of pronunciation features you control. A candidate with a strong regional accent can reach band 8 if they remain easy to follow.
What is actually assessed:
- Word and sentence stress. Stressing the wrong syllable causes more misunderstanding than any individual sound.
- Intonation. Flat delivery makes long answers hard to follow even when every word is correct.
- Chunking. Grouping words into meaningful phrases with pauses at the right places.
- Individual sounds, but only insofar as they affect understanding.
| Band 6 sounds like | Band 7 sounds like |
|---|---|
| Can generally be understood throughout, though mispronunciation of individual words or sounds reduces clarity at times. | Shows all the positive features of band 6 and some of band 8: a range of pronunciation features used with mixed control, generally easy to understand, with only occasional lapses. |
Habits that quietly cost half a band
- Short Part 1 answers. Treating the warm-up as small talk wastes five minutes of assessable speech and sets a low baseline.
- Stopping early in Part 2. Ending at 45 seconds because you have "finished" removes the very evidence the long turn exists to produce.
- Reciting a memorised answer. Detected, discounted, and it damages Fluency when the topic shifts and the script runs out.
- Over-correcting. Restarting a sentence three times to fix an article costs more in Fluency than the article cost in Grammar.
- Speaking too quietly. Anything the examiner cannot hear cannot be credited.
Because Speaking is assessed live and in real time, self-assessment is unreliable — most candidates cannot hear their own hesitation patterns. Recording an answer and reading a transcript against the four criteria is far more revealing than another practice conversation. The same four-criteria logic applies to Writing Task 2, where the fix is easier because you can see the evidence on the page.
Frequently asked
What are the four IELTS Speaking criteria?
Fluency and Coherence, Lexical Resource, Grammatical Range and Accuracy, and Pronunciation. Each is scored from 0 to 9 with equal weight, and your Speaking band is the average of the four.
Does having an accent lower your IELTS Speaking score?
No. Pronunciation is marked on how easy you are to understand, not on how closely you resemble a British or American speaker. A strong accent that remains clear can reach band 8; unclear speech in any accent cannot.
How long should you speak in IELTS Speaking Part 2?
Aim to fill the full two minutes. Stopping at 45 seconds limits what the examiner can assess and usually caps Fluency and Coherence, because sustained speech is exactly what this part is testing.
Is it bad to hesitate in IELTS Speaking?
It depends on why. Pausing to think about the content of an idea is normal and does not reduce your band. Pausing to search for the word or the grammar is what the descriptors count as hesitation, and repeated searching keeps Fluency at band 6.
Can you ask the examiner to repeat a question?
Yes, in Parts 1 and 3. Asking for repetition or clarification is natural spoken behaviour and is not penalised. The examiner cannot rephrase Part 2 cue card wording, however.